技术领域Technical Field
本发明涉及金融风险管理领域，尤其涉及一种多维动态评估的不动产抵押物金融风险防控方法与系统。The present invention relates to the field of financial risk management, and in particular to a multi-dimensional dynamic assessment method and system for preventing and controlling financial risks of real estate mortgages.
背景技术Background technique
在传统金融领域，不动产抵押贷款是一种常见的贷款方式，其核心是以不动产作为抵押物来确保贷款的还款。不动产的评估价值直接影响到贷款额度、利率以及风险控制策略的制定。然而，现有的不动产抵押物评估方法存在以下几方面的问题和不足：In the traditional financial field, real estate mortgage loans are a common form of loan, the core of which is to use real estate as collateral to ensure loan repayment. The assessed value of real estate directly affects the loan amount, interest rate, and the formulation of risk control strategies. However, the existing real estate collateral assessment methods have the following problems and shortcomings:
1.单一数据源和静态评估：传统评估常常依赖于历史交易数据和专家经验，缺乏对实时市场动态的考虑，无法反映市场短期内的波动和趋势。1. Single data source and static evaluation: Traditional evaluation often relies on historical transaction data and expert experience, lacks consideration of real-time market dynamics, and cannot reflect short-term market fluctuations and trends.
2.缺乏动态调整机制：一旦完成初步的风险评估后，很少有模型能够根据市场的实时变化进行动态调整。这导致评估结果很快过时，无法指导实时的风险防控。2. Lack of dynamic adjustment mechanism: Once the initial risk assessment is completed, few models can be dynamically adjusted according to real-time changes in the market. This causes the assessment results to become outdated quickly and cannot guide real-time risk prevention and control.
3.忽视宏观经济因素：除了不动产本身的因素外，宏观经济环境、政策导向、利率变动等都会影响不动产的价值和风险状况，而这些因素在传统模型中往往得不到充分的考虑。3. Ignoring macroeconomic factors: In addition to factors related to the real estate itself, the macroeconomic environment, policy orientation, interest rate changes, etc. will affect the value and risk status of real estate, and these factors are often not fully considered in traditional models.
4.评估效率低下：手工或半自动化的评估流程效率低下，无法满足大规模、快节奏的金融服务需求。4. Inefficient assessment: Manual or semi-automated assessment processes are inefficient and cannot meet the large-scale, fast-paced financial service needs.
5.风险预测能力弱：传统模型通常侧重于当前风险的评估，而不是未来风险的预测，这限制了金融机构在风险管理上的前瞻性。5. Weak risk prediction capabilities: Traditional models usually focus on the assessment of current risks rather than the prediction of future risks, which limits the foresight of financial institutions in risk management.
因此，现有技术亟需一种能够综合多源数据，具备动态调整和实时响应市场变化能力的不动产抵押物金融风险评估和防控模型，以提高风险评估的准确性和效率，减少金融系统的系统性风险。Therefore, existing technologies urgently need a real estate mortgage financial risk assessment and prevention model that can integrate multi-source data and has the ability to dynamically adjust and respond to market changes in real time, so as to improve the accuracy and efficiency of risk assessment and reduce the systemic risks of the financial system.
发明内容Summary of the invention
有鉴于此，本发明提供了一种多维动态评估的不动产抵押物金融风险防控方法与系统，以解决现有技术中不动产抵押物评估的静态性和延迟性问题，提高金融机构在不动产抵押贷款业务中的风险评估和管理效率。In view of this, the present invention provides a multi-dimensional dynamic assessment method and system for real estate mortgage financial risk prevention and control to solve the static and delayed problems in real estate mortgage assessment in the prior art and to improve the risk assessment and management efficiency of financial institutions in real estate mortgage loan business.
为实现上述目的，本发明提供如下技术方案：To achieve the above object, the present invention provides the following technical solutions:
第一方面，本发明提供了一种多维动态评估的不动产抵押物金融风险防控方法，包括以下步骤：In a first aspect, the present invention provides a multi-dimensional dynamic assessment method for real estate mortgage financial risk prevention and control, comprising the following steps:
收集与不动产抵押物贷款相关的多维数据，并进行数据清洗和预处理；Collect multi-dimensional data related to real estate mortgage loans, and perform data cleaning and pre-processing;
根据预处理后的多维数据，确定不动产抵押物的关键特征要素，并作为知识图谱中的节点；Based on the pre-processed multi-dimensional data, the key characteristic elements of the real estate mortgage are determined and used as nodes in the knowledge graph;
运用实体识别和关系抽取建立所述节点之间的关联关系，构建特征知识图谱，并应用图神经网络处理所述特征知识图谱，学习节点的独立特征及节点间的关联特征；Using entity recognition and relationship extraction to establish the association relationship between the nodes, constructing a feature knowledge graph, and applying a graph neural network to process the feature knowledge graph to learn the independent features of the nodes and the association features between the nodes;
利用特征知识图谱中提取的特征训练风险评估模型，并通过训练的风险评估模型预测不动产抵押物的未来价值，基于节点特征计算风险值，对风险值进行排序，形成风险评估模型的输出；The features extracted from the feature knowledge graph are used to train the risk assessment model, and the future value of the real estate mortgage is predicted through the trained risk assessment model. The risk value is calculated based on the node features, and the risk values are sorted to form the output of the risk assessment model.
将风险评估模型的输出与事先设定的风险阈值进行比较，一旦监测到任何风险值超过风险阈值时，自动触发预警机制，生成风险评估结果并启动风险防控措施，同时根据风险评估结果调整金融机构的抵押贷款政策。The output of the risk assessment model is compared with the pre-set risk threshold. Once any risk value is detected to exceed the risk threshold, the early warning mechanism is automatically triggered to generate risk assessment results and initiate risk prevention and control measures. At the same time, the mortgage loan policy of the financial institution is adjusted according to the risk assessment results.
作为本发明的进一步方案，收集的与不动产抵押物贷款相关的多维数据包括：不动产抵押物的市场数据、建筑法规数据、经济指标数据、不动产特征数据、借款人信息数据以及地理价值数据；其中：As a further solution of the present invention, the multi-dimensional data collected related to real estate mortgage loans include: market data of real estate mortgages, building regulations data, economic indicator data, real estate feature data, borrower information data and geographic value data; wherein:
所述市场数据包括市场价值变化数据、租金水平数据和供需情况数据，所述市场价值变化数据包括不动产抵押物的历史成交价、当前市场估价以及地区发展趋势数据；所述租金水平数据包括收集租金的历史数据和预测数据；所述供需情况数据包括区域内不动产抵押物的供应量数据和需求量数据；The market data includes market value change data, rental level data and supply and demand data. The market value change data includes historical transaction prices, current market valuations and regional development trend data of real estate mortgages; the rental level data includes historical data and forecast data of collected rents; the supply and demand data includes supply data and demand data of real estate mortgages in the region;
所述建筑法规数据包括土地使用权数据以及建筑规范数据，所述土地使用权数据包括土地使用年限、规划用途以及使用权转让限制数据；所述建筑规范数据包括建筑面积以及容积率数据；The building regulation data includes land use right data and building regulation data, wherein the land use right data includes land use period, planned use and land use right transfer restriction data; the building regulation data includes building area and volume ratio data;
所述经济指标数据包括宏观经济数据和行业发展数据；所述宏观经济数据包括GDP增长率、通货膨胀率和利率数据；The economic indicator data include macroeconomic data and industry development data; the macroeconomic data include GDP growth rate, inflation rate and interest rate data;
所述不动产特征数据包括抵押物物理特征数据和抵押物交易特征数据；所述抵押物物理特征数据包括抵押物的位置、面积、建筑年代、建筑物质量和设施配套数据；所述抵押物交易特征数据包括抵押物的历史交易记录和产权数据；The real estate characteristic data includes the physical characteristic data of the mortgage and the transaction characteristic data of the mortgage; the physical characteristic data of the mortgage includes the location, area, construction age, building quality and supporting facilities data of the mortgage; the transaction characteristic data of the mortgage includes the historical transaction records and property rights data of the mortgage;
所述借款人信息数据包括借款人的信用评分、历史逾期记录、收入数据、负债比和资产数据；The borrower information data includes the borrower's credit score, historical overdue records, income data, debt ratio and asset data;
所述地理价值数据包括抵押物的社区安全指数和生活便利性指数。The geographic value data includes the community safety index and life convenience index of the mortgaged property.
作为本发明的进一步方案，确定不动产抵押物的关键特征要素，包括以下步骤：As a further solution of the present invention, determining the key characteristic elements of the immovable property mortgage includes the following steps:
将收集到的多维数据导入Python的pandas库中，运行describe（）统计函数获取多维数据的总结统计结果，并使用直方图检查数据分布；Import the collected multidimensional data into Python’s pandas library, run the describe() statistical function to obtain summary statistics of the multidimensional data, and use a histogram to check the data distribution;
使用Python的pandas库中的corr（）函数计算皮尔逊相关系数，并使用Python的seaborn库中的heatmap（）函数，根据皮尔逊相关系数矩阵生成热图；Use the corr() function in Python's pandas library to calculate the Pearson correlation coefficient, and use the heatmap() function in Python's seaborn library to generate a heat map based on the Pearson correlation coefficient matrix;
使用ANOVA测试进行单变量分析，并根据总结统计结果和热图对多维数据分类，计算P值，选择P值低于预设阈值的特征；Univariate analysis was performed using ANOVA test, and multidimensional data were classified based on summary statistics and heat maps, P values were calculated, and features with P values below a preset threshold were selected;
使用递归特征消除（RFE）迭代选择特征，通过递归减少特征集确定最有影响力的特征作为不动产抵押物的关键特征要素；Iteratively select features using recursive feature elimination (RFE) to determine the most influential features as key feature elements of real estate collateral by recursively reducing the feature set;
使用k-fold交叉验证将多维数据分割成k组，进行k次训练和验证，每次选择不同的验证集和训练集，计算k次验证结果的平均准确度，对关键特征要素进行验证。Use k-fold cross validation to divide the multidimensional data into k groups, perform k training and validation, select different validation sets and training sets each time, calculate the average accuracy of the k validation results, and validate the key feature elements.
作为本发明的进一步方案，确定的不动产抵押物的关键特征要素包括：As a further solution of the present invention, the key characteristic elements of the real estate mortgage determined include:
物理特征，包含：地理位置、建筑面积、房间数、建筑年代、建筑材料以及楼层数；Physical characteristics, including: geographic location, floor area, number of rooms, construction age, building materials, and number of floors;
法律特征，包含：产权证明、土地使用权、规划用途、抵押信息以及历史交易记录；Legal features, including: property rights, land use rights, planned use, mortgage information, and historical transaction records;
价值特征，包含：估价值、税收评估、市场价值以及历史成交价；Value characteristics, including: estimated value, tax assessment, market value, and historical transaction price;
环境特征，包含：周边设施、公共交通、教育资源以及环境质量。Environmental characteristics, including: surrounding facilities, public transportation, educational resources and environmental quality.
作为本发明的进一步方案，关键特征要素作为知识图谱中的节点，运用实体识别和关系抽取建立所述节点之间的关联关系，构建特征知识图谱，包括以下步骤：As a further solution of the present invention, the key feature elements are used as nodes in the knowledge graph, and the association relationship between the nodes is established by using entity recognition and relationship extraction to construct a feature knowledge graph, including the following steps:
选择图数据库作为知识图谱存储，加载预训练的NLP模型，并连接到图数据库；Select a graph database as the knowledge graph storage, load the pre-trained NLP model, and connect to the graph database;
定义知识图谱中所包含的实体类型，并运用NLP识别多维数据文本中的实体；Define the entity types contained in the knowledge graph and use NLP to identify entities in multidimensional data text;
定义实体之间关系类型，并运用NLP关系抽取识别实体之间的关联关系；Define the relationship types between entities and use NLP relationship extraction to identify the associations between entities;
合并不同来源以及格式中指代同一实体的多个实体表示，使实体关联关系对齐；Merge multiple entity representations referring to the same entity from different sources and formats to align entity associations;
将识别的实体和关联关系转化为图谱中的节点和边，并为知识图谱的节点和关系建立图谱索引。The identified entities and relationships are converted into nodes and edges in the graph, and a graph index is established for the nodes and relationships of the knowledge graph.
作为本发明的进一步方案，应用图神经网络处理所述特征知识图谱，学习节点的独立特征及节点间的关联特征，包括以下步骤：As a further solution of the present invention, a graph neural network is applied to process the feature knowledge graph to learn the independent features of nodes and the associated features between nodes, including the following steps:
基于特征知识图谱构造节点特征和边特征的特征向量，并选择GAT作为图神经网络架构；Construct feature vectors of node features and edge features based on the feature knowledge graph, and select GAT as the graph neural network architecture;
将图谱数据划分为训练集、验证集和测试集，定义损失函数，选择Adam优化器通过训练集进行图神经网络模型训练，通过图神经网络模型学习节点的独立特征和节点间的关联特征，通过验证集和测试集对图神经网络模型进行验证与调优；Divide the graph data into training set, validation set and test set, define the loss function, select Adam optimizer to train the graph neural network model through the training set, learn the independent features of nodes and the correlation features between nodes through the graph neural network model, and verify and tune the graph neural network model through the validation set and test set;
利用训练好的图神经网络模型对不动产抵押物的风险进行评估。Use the trained graph neural network model to assess the risk of real estate mortgages.
作为本发明的进一步方案，所述图神经网络模型学习节点的独立特征和节点间的关联特征时，通过GNN层传播节点特征，每个节点聚合邻居节点信息，使用聚合的邻居节点信息更新每个节点的特征，并进行池化操作。As a further solution of the present invention, when the graph neural network model learns the independent features of nodes and the correlation features between nodes, the node features are propagated through the GNN layer, each node aggregates the neighbor node information, the aggregated neighbor node information is used to update the features of each node, and a pooling operation is performed.
作为本发明的进一步方案，利用特征知识图谱中提取的特征训练风险评估模型，并通过训练的风险评估模型预测不动产抵押物的未来价值时，使用图神经网络模型作为特征提取器，提取的特征作为风险评估模型的输入；定义一个风险评估的对数损失函数，使用训练集数据训练风险评估模型，利用训练好的风险评估模型计算不动产抵押物的风险值，所述不动产抵押物的风险值包括违约概率和损失预期值，并基于市场趋势和经济指标预测不动产抵押物的未来价值，并根据计算出的风险值对不动产抵押物进行排序，形成风险评级。As a further solution of the present invention, a risk assessment model is trained using features extracted from a feature knowledge graph, and when the future value of a real estate mortgage is predicted through the trained risk assessment model, a graph neural network model is used as a feature extractor, and the extracted features are used as input to the risk assessment model; a logarithmic loss function for risk assessment is defined, the risk assessment model is trained using training set data, and the risk value of the real estate mortgage is calculated using the trained risk assessment model, wherein the risk value of the real estate mortgage includes a probability of default and an expected value of loss, and the future value of the real estate mortgage is predicted based on market trends and economic indicators, and the real estate mortgage is sorted according to the calculated risk value to form a risk rating.
作为本发明的进一步方案，计算不动产抵押物的风险值，包括以下步骤：As a further solution of the present invention, calculating the risk value of the real estate mortgage includes the following steps:
步骤1、特征提取：通过图神经网络从知识图谱中提取出每个不动产抵押物节点的特征向量；设fi表示第i个不动产抵押物的特征向量；Step 1, feature extraction: extract the feature vector of each real estate mortgage node from the knowledge graph through the graph neural network; letfi represent the feature vector of the i-th real estate mortgage;
步骤2、构建评估模型：使用逻辑回归分类模型对违约概率（DP）进行评估，使用线性回归模型对损失预期值（LGD）进行评估；Step 2: Construct an evaluation model: Use a logistic regression classification model to evaluate the probability of default (DP), and use a linear regression model to evaluate the loss expectancy (LGD);
步骤3、模型训练：使用训练集数据训练违约概率（DP）模型和损失预期值（LGD）模型，分别预测违约概率和损失预期值；Step 3: Model training: Use the training set data to train the default probability (DP) model and the loss expected value (LGD) model to predict the default probability and loss expected value respectively;
步骤4、计算风险值：通过违约概率和损失预期值的组合来计算风险值（RiskValue,RV）；Step 4: Calculate the risk value: Calculate the risk value (RiskValue, RV) by combining the probability of default and the expected value of loss;
步骤5、风险值排序：根据计算出的风险值，对所有不动产抵押物进行排序，风险值高的排在前面，风险值低的排在后面，形成风险评估模型的输出。Step 5: Risk value sorting: All real estate mortgages are sorted according to the calculated risk values, with those with high risk values at the front and those with low risk values at the back, forming the output of the risk assessment model.
作为本发明的进一步方案，违约概率（DP）模型表示为：As a further aspect of the present invention, the default probability (DP) model is expressed as:
； ;
式中，σ为sigmoid函数，用于将线性输出映射到概率空间（0到1之间）；Where σ is the sigmoid function, which is used to map the linear output to the probability space (between 0 and 1);
W为违约概率模型权重；fi第i个不动产抵押物的特征向量；b为偏置项；W is the weight of the default probability model; fiis the feature vector of the ith real estate mortgage; b is the bias term;
其中，损失预期值（LGD）模型表示为：Among them, the loss expected value (LGD) model is expressed as:
； ;
式中，W和b为通过最小化损失函数学习的模型参数。Where W and b are the model parameters learned by minimizing the loss function.
作为本发明的进一步方案，所述通过违约概率和损失预期值的组合来计算风险值，计算公式如下：As a further solution of the present invention, the risk value is calculated by combining the probability of default and the expected value of loss, and the calculation formula is as follows:
； ;
式中，RVi是第i个不动产抵押物的总体风险值。Where RVi is the overall risk value of the i-th real estate mortgage.
第二方面，本发明还提供了一种多维动态评估的不动产抵押物金融风险防控系统，包括以下模块：In a second aspect, the present invention also provides a multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control system, including the following modules:
数据收集模块，用于收集与不动产抵押物贷款相关的多维数据，并进行数据清洗和预处理；Data collection module, used to collect multi-dimensional data related to real estate mortgage loans, and perform data cleaning and pre-processing;
特征工程模块，用于根据预处理后的多维数据，确定不动产抵押物的关键特征要素，并作为知识图谱中的节点；The feature engineering module is used to determine the key feature elements of real estate collateral based on the pre-processed multi-dimensional data and use them as nodes in the knowledge graph;
知识图谱构建模块，用于运用实体识别和关系抽取建立所述节点之间的关联关系，构建特征知识图谱；A knowledge graph construction module is used to establish the association relationship between the nodes by using entity recognition and relationship extraction to construct a feature knowledge graph;
图神经网络处理模块，用于应用图神经网络处理所述特征知识图谱，学习节点的独立特征及节点间的关联特征；A graph neural network processing module, used to apply a graph neural network to process the feature knowledge graph, and learn independent features of nodes and associated features between nodes;
风险评估模块，用于利用特征知识图谱中提取的特征训练风险评估模型，并通过训练的风险评估模型预测不动产抵押物的未来价值，基于节点特征计算风险值，对风险值进行排序，形成风险评估模型的输出；The risk assessment module is used to train the risk assessment model using the features extracted from the feature knowledge graph, and predict the future value of the real estate mortgage through the trained risk assessment model, calculate the risk value based on the node features, sort the risk values, and form the output of the risk assessment model;
风险预警防控模块，用于将风险评估模型的输出与事先设定的风险阈值进行比较，一旦监测到任何风险值超过风险阈值时，自动触发预警机制，生成风险评估结果并启动风险防控措施，同时根据风险评估结果调整金融机构的抵押贷款政策。The risk early warning and control module is used to compare the output of the risk assessment model with the pre-set risk threshold. Once any risk value is detected to exceed the risk threshold, the early warning mechanism is automatically triggered to generate risk assessment results and initiate risk prevention and control measures. At the same time, the mortgage loan policy of the financial institution is adjusted according to the risk assessment results.
作为本发明的进一步方案，所述特征工程模块中包括：As a further solution of the present invention, the feature engineering module includes:
统计分析单元，用于运行describe（）等函数进行数据总结统计；Statistical analysis unit, used to run describe() and other functions to summarize data;
相关性分析单元，用于计算皮尔逊相关系数并生成热图；Correlation analysis unit, used to calculate Pearson correlation coefficient and generate heat map;
单变量分析单元，用于通过ANOVA测试筛选显著特征；Univariate analysis unit, used to screen significant features through ANOVA test;
特征选择单元，用于识别关键特征；A feature selection unit, used to identify key features;
交叉验证单元，用于通过k-fold交叉验证特征的稳定性和预测能力。The cross-validation unit is used to verify the stability and predictive power of features through k-fold cross-validation.
作为本发明的进一步方案，所述知识图谱构建模块包括：As a further solution of the present invention, the knowledge graph construction module includes:
图数据库，用于存储和管理知识图谱数据。Graph database, used to store and manage knowledge graph data.
NLP模型，用于实体识别和关系抽取。NLP models for entity recognition and relation extraction.
实体关联对齐单元，用于合并不同来源的同一实体表示。Entity association alignment unit, used to merge the same entity representations from different sources.
知识图谱索引构建单元，用于提高查询效率。Knowledge graph index building unit, used to improve query efficiency.
作为本发明的进一步方案，所述风险评估模块中采用图神经网络作为特征提取器，用于从知识图谱提取风险特征；风险评估的对数损失函数定义，用于指导模型训练；违约概率模型和损失预期值模型用于风险值评估。As a further solution of the present invention, a graph neural network is used as a feature extractor in the risk assessment module to extract risk features from the knowledge graph; the logarithmic loss function of risk assessment is defined to guide model training; the default probability model and the loss expectation value model are used for risk value assessment.
作为本发明的进一步方案，所述风险预警防控模块中包括：As a further solution of the present invention, the risk warning and control module includes:
风险阈值设定单元，用于设定风险预警阈值；A risk threshold setting unit, used to set a risk warning threshold;
风险评估结果处理单元，用于风险值排序和输出；以及A risk assessment result processing unit for ranking and outputting risk values; and
预警机制触发器，用于在风险值超过阈值时发出预警并采取相应措施。The early warning mechanism trigger is used to issue an early warning and take corresponding measures when the risk value exceeds the threshold.
作为本发明的进一步方案，所述多维动态评估的不动产抵押物金融风险防控系统，还包括策略调整模块，用于根据风险评估结果调整金融机构策略。As a further solution of the present invention, the multi-dimensional dynamic assessment system for preventing and controlling financial risks of real estate mortgages also includes a strategy adjustment module for adjusting the strategy of financial institutions according to risk assessment results.
与现有技术相比，本发明的多维动态评估的不动产抵押物金融风险防控方法及系统，具有如下有益效果：Compared with the prior art, the multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method and system of the present invention has the following beneficial effects:
1.提高了风险识别的准确性：1. Improved the accuracy of risk identification:
本发明利用知识图谱和图神经网络的结合，可以更准确地识别风险因素及其之间的复杂关系，提供了一个深度学习的环境，可以捕捉到传统统计方法可能忽略的非线性模式。The present invention utilizes the combination of knowledge graph and graph neural network to more accurately identify risk factors and the complex relationships between them, and provides a deep learning environment that can capture nonlinear patterns that may be overlooked by traditional statistical methods.
2.动态风险评估：2. Dynamic risk assessment:
传统的风险评估方法通常是静态的，而本发明能够实现动态评估，考虑市场和环境因素的实时变化，以及时间序列上的风险动态演化。Traditional risk assessment methods are usually static, while the present invention can achieve dynamic assessment, taking into account the real-time changes in market and environmental factors, as well as the dynamic evolution of risks over time series.
3.实时风险监控和预警：3. Real-time risk monitoring and early warning:
本发明的系统设计有风险预警防控模块，能够在风险发生前发出警告，并采取预防措施，从而最小化潜在的金融损失。The system of the present invention is designed with a risk warning and prevention module, which can issue warnings before risks occur and take preventive measures to minimize potential financial losses.
4.政策调整的灵活性：4. Flexibility of policy adjustments:
随着风险评估模型的持续运作，金融机构能够根据模型提供的最新数据和分析结果，动态调整贷款政策和风险控制策略，增强应对市场波动的能力。自动化和智能化的风险评估减少了人工的干预，降低了运营成本和错误率。此外，提前的风险防控能降低潜在的违约和损失成本。With the continuous operation of the risk assessment model, financial institutions can dynamically adjust loan policies and risk control strategies based on the latest data and analysis results provided by the model, and enhance their ability to cope with market fluctuations. Automated and intelligent risk assessment reduces manual intervention, reduces operating costs and error rates. In addition, early risk prevention and control can reduce potential default and loss costs.
5.增强策略制定的数据支持：5. Enhance data support for strategy formulation:
本发明的系统提供的数据分析和报告为制定金融产品和服务提供了数据支撑，帮助金融机构更好地设计产品，满足市场需求。本发明的系统设计允许持续学习和优化，通过不断的模型训练和调整，系统的预测能力和风险控制策略将随着时间的推移而持续改进。The data analysis and reports provided by the system of the present invention provide data support for the formulation of financial products and services, helping financial institutions to better design products and meet market demand. The system design of the present invention allows for continuous learning and optimization. Through continuous model training and adjustment, the system's predictive capabilities and risk control strategies will continue to improve over time.
综上所述，本发明不仅提升了风险评估与管理的效率和准确性，同时也为金融机构提供了一个高度自动化、智能化的风险防控工具。本发明的系统有助于实现更加稳健的财务管理，并且对于防范和控制金融风险具有重要的实用价值。In summary, the present invention not only improves the efficiency and accuracy of risk assessment and management, but also provides a highly automated and intelligent risk prevention and control tool for financial institutions. The system of the present invention helps to achieve more robust financial management and has important practical value for preventing and controlling financial risks.
本发明的这些方面或其他方面在以下实施例的描述中会更加简明易懂。应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本发明。These and other aspects of the present invention will become more concise and understandable in the following description of the embodiments. It should be understood that the above general description and the following detailed description are only exemplary and explanatory and cannot limit the present invention.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
为了更清楚地说明本发明实施例或相关技术中的技术方案，下面将对示例性实施例或相关技术描述中所需要使用的附图作一简单地介绍，附图用来提供对本发明的进一步理解，并且构成说明书的一部分，与本发明的实施例一起用于解释本发明，并不构成对本发明的限制。在附图中：In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the following briefly introduces the drawings required for use in the exemplary embodiments or related technical descriptions. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:
图1为本发明实施例的一种多维动态评估的不动产抵押物金融风险防控方法的流程图。FIG1 is a flow chart of a multi-dimensional dynamic assessment method for preventing and controlling financial risks of real estate mortgages according to an embodiment of the present invention.
图2为本发明实施例的一种多维动态评估的不动产抵押物金融风险防控方法中确定不动产抵押物的关键特征要素的流程图。2 is a flow chart of determining key characteristic elements of real estate mortgages in a multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method according to an embodiment of the present invention.
图3为本发明实施例的一种多维动态评估的不动产抵押物金融风险防控方法中构建特征知识图谱的流程图。FIG3 is a flow chart of constructing a feature knowledge graph in a multi-dimensional dynamic assessment method for preventing and controlling financial risks of real estate mortgages according to an embodiment of the present invention.
图4为本发明实施例的一种多维动态评估的不动产抵押物金融风险防控方法中计算不动产抵押物的风险值的流程图。4 is a flow chart of calculating the risk value of real estate mortgage in a multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method according to an embodiment of the present invention.
具体实施方式Detailed ways
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本发明，并不用于限定本发明。In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
在本发明的说明书和权利要求书及上述附图中的描述的一些流程中，包含了按照特定顺序出现的多个操作，但是应该清楚了解，这些操作可以不按照其在本文中出现的顺序来执行或并行执行，操作的序号如101、102等，仅仅是用于区分开各个不同的操作，序号本身不代表任何的执行顺序。另外，这些流程可以包括更多或更少的操作，并且这些操作可以按顺序执行或并行执行。需要说明的是，本文中的“第一”、“第二”等描述，是用于区分不同的消息、设备、模块等，不代表先后顺序，也不限定“第一”和“第二”是不同的类型。In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.
下面将结合本发明示例性实施例中的附图，对本发明示例性实施例中的技术方案进行清楚、完整地描述，显然，所描述的示例性实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The following will be combined with the drawings in the exemplary embodiments of the present invention to clearly and completely describe the technical solutions in the exemplary embodiments of the present invention. Obviously, the exemplary embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
在不动产抵押贷款中，以不动产作为抵押物来确保贷款的还款，不动产的评估价值直接影响到贷款额度、利率以及风险控制策略的制定。现有的不动产抵押物评估方法存在不动产抵押物评估的静态性和延迟性问题。为解决上述问题，本发明提供了一种多维动态评估的不动产抵押物金融风险防控方法与系统，利用知识图谱和图神经网络的结合，可以更准确地识别风险因素及其之间的复杂关系，提供了一个深度学习的环境，可以捕捉到传统统计方法可能忽略的非线性模式。本发明能够实现动态评估，考虑市场和环境因素的实时变化，以及时间序列上的风险动态演化，有助于实现更加稳健的财务管理，并且对于防范和控制金融风险具有重要的实用价值。In real estate mortgage loans, real estate is used as collateral to ensure loan repayment, and the assessed value of real estate directly affects the loan amount, interest rate, and the formulation of risk control strategies. Existing real estate mortgage assessment methods have static and delayed real estate mortgage assessment problems. To solve the above problems, the present invention provides a multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method and system, which utilizes the combination of knowledge graph and graph neural network to more accurately identify risk factors and the complex relationships between them, and provides a deep learning environment that can capture nonlinear patterns that traditional statistical methods may ignore. The present invention can achieve dynamic assessment, taking into account real-time changes in market and environmental factors, as well as the dynamic evolution of risks in time series, which helps to achieve more robust financial management, and has important practical value for preventing and controlling financial risks.
下面结合具体实施例对本发明的技术方案作进一步的说明：The technical solution of the present invention is further described below in conjunction with specific embodiments:
参见图1所示，本发明实施例提供的一种多维动态评估的不动产抵押物金融风险防控方法，包括如下步骤：As shown in FIG1 , a multi-dimensional dynamic assessment method for real estate mortgage financial risk prevention and control provided by an embodiment of the present invention includes the following steps:
步骤1）、收集与不动产抵押物贷款相关的多维数据，并进行数据清洗和预处理；Step 1) Collect multi-dimensional data related to real estate mortgage loans, and perform data cleaning and preprocessing;
步骤2）、根据预处理后的多维数据，确定不动产抵押物的关键特征要素，并作为知识图谱中的节点；Step 2) According to the pre-processed multi-dimensional data, determine the key characteristic elements of the real estate mortgage and use them as nodes in the knowledge graph;
步骤3）、运用实体识别和关系抽取建立所述节点之间的关联关系，构建特征知识图谱，并应用图神经网络处理所述特征知识图谱，学习节点的独立特征及节点间的关联特征；Step 3), using entity recognition and relationship extraction to establish the association relationship between the nodes, constructing a feature knowledge graph, and applying a graph neural network to process the feature knowledge graph to learn the independent features of the nodes and the association features between the nodes;
步骤4）、利用特征知识图谱中提取的特征训练风险评估模型，并通过训练的风险评估模型预测不动产抵押物的未来价值，基于节点特征计算风险值，对风险值进行排序，形成风险评估模型的输出；Step 4) Use the features extracted from the feature knowledge graph to train the risk assessment model, and use the trained risk assessment model to predict the future value of the real estate mortgage, calculate the risk value based on the node features, sort the risk values, and form the output of the risk assessment model;
步骤5）、将风险评估模型的输出与事先设定的风险阈值进行比较，一旦监测到任何风险值超过风险阈值时，自动触发预警机制，生成风险评估结果并启动风险防控措施，同时根据风险评估结果调整金融机构的抵押贷款政策。Step 5) Compare the output of the risk assessment model with the pre-set risk threshold. Once any risk value is detected to exceed the risk threshold, the early warning mechanism is automatically triggered to generate risk assessment results and initiate risk prevention and control measures. At the same time, the mortgage loan policy of the financial institution is adjusted according to the risk assessment results.
示例性的，以一家金融机构为例，在使用该多维动态评估的不动产抵押物金融风险防控方法进行不动产抵押物金融风险防控时，首先，金融机构从各种来源收集与不动产抵押物贷款相关的多维数据，包括市场数据、建筑法规数据、经济指标数据、不动产特征数据、借款人信息数据和地理价值数据。他们对数据进行清洗和预处理，确保数据的准确性和完整性。然后，确定关键特征要素，通过数据分析和专业知识，金融机构确定了关键特征要素，如市场价值变化数据、土地使用权数据、GDP增长率等，这些特征将在后续的风险评估中起到重要作用。然后，实体识别和关系抽取，利用自然语言处理和机器学习技术，金融机构建立了特征知识图谱，揭示了不动产抵押物及其相关要素之间的关联关系，为风险评估提供了更深入的理解。然后对风险评估模型训练，基于特征知识图谱提取的特征，金融机构训练风险评估模型，该模型能够预测不动产抵押物的未来价值并计算风险值，帮助他们更好地评估贷款风险。最后，进行风险阈值设定和预警机制，金融机构设定了风险阈值，并与风险评估模型的输出进行比较。一旦某个不动产抵押物的风险值超过设定的阈值，系统将自动触发预警机制，通知相关人员进行进一步审查和风险控制措施的制定。For example, taking a financial institution as an example, when using the multi-dimensional dynamic evaluation of real estate mortgage financial risk prevention and control method for real estate mortgage financial risk prevention and control, first, the financial institution collects multi-dimensional data related to real estate mortgage loans from various sources, including market data, building regulations data, economic indicator data, real estate feature data, borrower information data and geographic value data. They clean and pre-process the data to ensure the accuracy and completeness of the data. Then, determine the key feature elements. Through data analysis and professional knowledge, the financial institution determines the key feature elements, such as market value change data, land use rights data, GDP growth rate, etc. These features will play an important role in subsequent risk assessment. Then, entity recognition and relationship extraction, using natural language processing and machine learning technology, the financial institution establishes a feature knowledge graph, which reveals the correlation between real estate mortgages and their related elements, providing a deeper understanding for risk assessment. Then, the risk assessment model is trained. Based on the features extracted from the feature knowledge graph, the financial institution trains a risk assessment model, which can predict the future value of real estate mortgages and calculate the risk value, helping them to better assess loan risks. Finally, risk threshold setting and early warning mechanism are carried out. The financial institution sets the risk threshold and compares it with the output of the risk assessment model. Once the risk value of a real estate mortgage exceeds the set threshold, the system will automatically trigger the early warning mechanism and notify relevant personnel to conduct further review and formulate risk control measures.
通过实施上述的多维动态评估的不动产抵押物金融风险防控方法，金融机构能够更准确、及时地评估不动产抵押物的价值和风险，从而有效降低不良贷款的风险，保障金融机构的资产安全和稳健经营。By implementing the above-mentioned multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control methods, financial institutions can more accurately and timely assess the value and risks of real estate mortgages, thereby effectively reducing the risk of non-performing loans and ensuring the asset security and sound operation of financial institutions.
在本发明中，步骤1）中收集的与不动产抵押物贷款相关的多维数据包括：不动产抵押物的市场数据、建筑法规数据、经济指标数据、不动产特征数据、借款人信息数据以及地理价值数据。In the present invention, the multi-dimensional data related to real estate mortgage loans collected in step 1) include: market data of real estate mortgages, building regulations data, economic indicator data, real estate feature data, borrower information data and geographic value data.
其中，所述市场数据包括市场价值变化数据、租金水平数据和供需情况数据，所述市场价值变化数据包括不动产抵押物的历史成交价、当前市场估价以及地区发展趋势数据；所述租金水平数据包括收集租金的历史数据和预测数据；所述供需情况数据包括区域内不动产抵押物的供应量数据和需求量数据；The market data includes market value change data, rental level data and supply and demand data. The market value change data includes historical transaction prices, current market valuations and regional development trend data of real estate mortgages; the rental level data includes historical data and forecast data of collected rents; the supply and demand data includes supply and demand data of real estate mortgages in the region;
所述建筑法规数据包括土地使用权数据以及建筑规范数据，所述土地使用权数据包括土地使用年限、规划用途以及使用权转让限制数据；所述建筑规范数据包括建筑面积以及容积率数据；The building regulation data includes land use right data and building regulation data, wherein the land use right data includes land use period, planned use and land use right transfer restriction data; the building regulation data includes building area and volume ratio data;
所述经济指标数据包括宏观经济数据和行业发展数据；所述宏观经济数据包括GDP增长率、通货膨胀率和利率数据；The economic indicator data include macroeconomic data and industry development data; the macroeconomic data include GDP growth rate, inflation rate and interest rate data;
所述不动产特征数据包括抵押物物理特征数据和抵押物交易特征数据；所述抵押物物理特征数据包括抵押物的位置、面积、建筑年代、建筑物质量和设施配套数据；所述抵押物交易特征数据包括抵押物的历史交易记录和产权数据；The real estate characteristic data includes the physical characteristic data of the mortgage and the transaction characteristic data of the mortgage; the physical characteristic data of the mortgage includes the location, area, construction age, building quality and supporting facilities data of the mortgage; the transaction characteristic data of the mortgage includes the historical transaction records and property rights data of the mortgage;
所述借款人信息数据包括借款人的信用评分、历史逾期记录、收入数据、负债比和资产数据；The borrower information data includes the borrower's credit score, historical overdue records, income data, debt ratio and asset data;
所述地理价值数据包括抵押物的社区安全指数和生活便利性指数。The geographic value data includes the community safety index and life convenience index of the mortgaged property.
除上述数据外，还可以根据不动产抵押物所处的环境因素数据，如周边环境污染情况、自然灾害风险等，对不动产价值和风险也有影响。以及考虑市场趋势数据，根据市场趋势数据了解不动产市场的发展趋势和预测未来走势，有助于更准确地评估抵押物价值和风险。In addition to the above data, the environmental factors data of the real estate mortgage, such as the surrounding environmental pollution, natural disaster risks, etc., can also affect the value and risk of real estate. As well as considering market trend data, understanding the development trend of the real estate market and predicting future trends based on market trend data will help to more accurately assess the value and risk of the mortgage.
参见图2所示，在本发明中，步骤2）中确定不动产抵押物的关键特征要素，包括以下步骤：As shown in FIG. 2 , in the present invention, determining the key characteristic elements of the immovable property mortgage in step 2) includes the following steps:
a）将收集到的多维数据导入Python的pandas库中，运行describe（）统计函数获取多维数据的总结统计结果，并使用直方图检查数据分布。a) Import the collected multidimensional data into Python’s pandas library, run the describe() statistical function to obtain summary statistics of the multidimensional data, and use a histogram to examine the data distribution.
在该步骤中，可以将收集到的市场数据、建筑法规数据、经济指标数据、不动产特征数据、借款人信息数据和地理价值数据导入Python的pandas库中，然后利用describe（）函数得出各个特征的统计摘要，并通过直方图检查数据的分布情况。In this step, the collected market data, building regulations data, economic indicator data, real estate characteristics data, borrower information data, and geographic value data can be imported into Python’s pandas library, and then the describe() function can be used to obtain a statistical summary of each feature, and the distribution of the data can be checked through a histogram.
其中，利用describe（）函数得出各个特征的统计摘要时，可以按照以下示例性的详细步骤操作：When using the describe() function to obtain the statistical summary of each feature, you can follow the following exemplary detailed steps:
（1）导入必要的库：(1) Import necessary libraries:
import pandas as pdimport pandas as pd
（2）创建数据框（DataFrame）：(2) Create a DataFrame:
# 创建一个示例数据框# Create a sample data frame
data={'Feature1':[10,20,30,40,50],data={'Feature1':[10,20,30,40,50],
'Feature2':[1.5,2.5,3.5,4.5,5.5],'Feature2':[1.5,2.5,3.5,4.5,5.5],
'Feature3':['A','B','C','D','E']}'Feature3':['A','B','C','D','E']}
df=pd.DataFrame(data)df = pd.DataFrame(data)
（3）使用`describe()`函数生成统计摘要：(3) Use the describe() function to generate a statistical summary:
summary=df.describe()summary = df.describe()
print(summary)print(summary)
输出结果将会包含均值、标准差、最小值、最大值、中位数的特征的统计摘要。The output will contain a statistical summary of the features including mean, standard deviation, minimum, maximum, and median.
（4）若要查看某一列的统计摘要，使用以下方式：(4) To view the statistical summary of a column, use the following method:
# 查看特定特征列的统计摘要# View statistical summary of a specific feature column
summary_feature1=df['Feature1'].describe()summary_feature1 = df['Feature1'].describe()
print(summary_feature1)print(summary_feature1)
在以上示例中，通过创建一个简单的数据框，包含三个特征。通过调用describe（）函数，可以得出这些特征的统计摘要，快速了解数据的分布情况。In the above example, a simple data frame is created with three features. By calling the describe() function, a statistical summary of these features can be obtained to quickly understand the distribution of the data.
b）使用Python的pandas库中的corr（）函数计算皮尔逊相关系数，并使用Python的seaborn库中的heatmap（）函数，根据皮尔逊相关系数矩阵生成热图。b) Use the corr() function in Python’s pandas library to calculate the Pearson correlation coefficient and use the heatmap() function in Python’s seaborn library to generate a heatmap based on the Pearson correlation coefficient matrix.
在该步骤中，使用corr（）函数计算各个特征之间的皮尔逊相关系数，通过seaborn库的heatmap（）函数生成热图，直观展现不同特征之间的相关性强弱，帮助确定关键特征。In this step, the corr() function is used to calculate the Pearson correlation coefficient between each feature, and the heatmap() function of the seaborn library is used to generate a heat map to intuitively show the strength of the correlation between different features and help identify key features.
c）使用ANOVA测试进行单变量分析，并根据总结统计结果和热图对多维数据分类，计算P值，选择P值低于预设阈值的特征。c) Perform univariate analysis using ANOVA test, and classify multidimensional data based on summary statistics and heat maps, calculate P values, and select features with P values below a preset threshold.
d）使用递归特征消除（RFE）迭代选择特征，通过递归减少特征集确定最有影响力的特征作为不动产抵押物的关键特征要素。d) Recursive feature elimination (RFE) is used to iteratively select features, and the most influential features are determined as key feature elements of real estate collateral by recursively reducing the feature set.
e）使用k-fold交叉验证将多维数据分割成k组，进行k次训练和验证，每次选择不同的验证集和训练集，计算k次验证结果的平均准确度，对关键特征要素进行验证。e) Use k-fold cross validation to divide the multidimensional data into k groups, perform k training and validation, select different validation sets and training sets each time, calculate the average accuracy of the k validation results, and validate the key feature elements.
例如，对于房地产抵押贷款，本发明可以收集包括市场价值、土地规划、借款人信息、建筑特征等多维数据，通过Python的pandas库和seaborn库进行数据导入、统计分析和相关性分析，识别出与抵押物价值相关的关键特征。然后利用ANOVA测试和递归特征消除方法，筛选出对抵押物价值有重要影响的特征。最后，通过k-fold交叉验证选定的关键特征对抵押物价值的预测准确度，并评估模型的稳定性和泛化能力。For example, for real estate mortgage loans, the present invention can collect multidimensional data including market value, land planning, borrower information, building features, etc., and use Python's pandas library and seaborn library to import data, perform statistical analysis and correlation analysis, and identify key features related to the value of the collateral. Then, the ANOVA test and recursive feature elimination method are used to screen out features that have a significant impact on the value of the collateral. Finally, the prediction accuracy of the selected key features for the value of the collateral is verified by k-fold cross-validation, and the stability and generalization ability of the model are evaluated.
本发明中结合了数据处理、统计分析、特征选择和模型验证等多个步骤，可以帮助金融机构更准确地评估不动产抵押贷款的风险和价值。也可以深入理解不动产抵押物的关键特征要素，提高贷款审批和风险控制的效率和准确性。通过以上步骤的实施，金融机构可以更好地理解和评估不动产抵押贷款的价值和风险，提高决策的科学性和准确性，从而实现更加稳健的贷款业务运营。The present invention combines multiple steps such as data processing, statistical analysis, feature selection and model verification, which can help financial institutions more accurately assess the risks and value of real estate mortgage loans. It can also deeply understand the key characteristic elements of real estate collateral and improve the efficiency and accuracy of loan approval and risk control. Through the implementation of the above steps, financial institutions can better understand and assess the value and risks of real estate mortgage loans, improve the scientificity and accuracy of decision-making, and thus achieve more robust loan business operations.
其中，确定的不动产抵押物的关键特征要素包括物理特征、法律特征、价值特征以及环境特征。其中，物理特征包含地理位置、建筑面积、房间数、建筑年代、建筑材料以及楼层数等。法律特征包含产权证明、土地使用权、规划用途、抵押信息以及历史交易记录等。价值特征包含估价值、税收评估、市场价值以及历史成交价。环境特征包含周边设施、公共交通、教育资源以及环境质量。Among them, the key characteristic elements of the real estate mortgage include physical characteristics, legal characteristics, value characteristics and environmental characteristics. Among them, physical characteristics include geographical location, building area, number of rooms, construction age, building materials and number of floors. Legal characteristics include property rights, land use rights, planned use, mortgage information and historical transaction records. Value characteristics include estimated value, tax assessment, market value and historical transaction price. Environmental characteristics include surrounding facilities, public transportation, educational resources and environmental quality.
参见图3所示，在步骤3）中，关键特征要素作为知识图谱中的节点，运用实体识别和关系抽取建立所述节点之间的关联关系，构建特征知识图谱，包括以下步骤：As shown in FIG. 3 , in step 3), the key feature elements are used as nodes in the knowledge graph, and entity recognition and relationship extraction are used to establish the association relationship between the nodes to construct the feature knowledge graph, including the following steps:
步骤S301、选择图数据库作为知识图谱存储，加载预训练的NLP模型，并连接到图数据库。Step S301: Select a graph database as the knowledge graph storage, load the pre-trained NLP model, and connect to the graph database.
步骤S302、定义知识图谱中所包含的实体类型，并运用NLP识别多维数据文本中的实体；Step S302: define the entity types contained in the knowledge graph, and use NLP to identify entities in the multidimensional data text;
步骤S303、定义实体之间关系类型，并运用NLP关系抽取识别实体之间的关联关系；Step S303: define the relationship type between entities, and use NLP relationship extraction to identify the association relationship between entities;
步骤S304、合并不同来源以及格式中指代同一实体的多个实体表示，使实体关联关系对齐；Step S304: merging multiple entity representations referring to the same entity in different sources and formats to align entity association relationships;
步骤S305、将识别的实体和关联关系转化为图谱中的节点和边，并为知识图谱的节点和关系建立图谱索引。Step S305: Convert the identified entities and relationships into nodes and edges in the graph, and establish a graph index for the nodes and relationships of the knowledge graph.
在上述步骤中，选择Neo4j作为图数据库，用于存储特征知识图谱，通过NLP技术识别出不同数据文本中的特征名称和关联统计指标，建立特征之间的关联关系，转化为图谱中的节点和边，构建特征知识图谱的网络结构。本发明中利用图数据库、NLP实体识别和关系抽取技术，构建特征知识图谱，将特征之间的关联关系以图谱的形式呈现，有助于更好地理解特征之间的关系，为数据分析、特征工程和机器学习模型的建立提供更深入的洞察和支持。同时，通过建立图谱索引，可以提高图谱的查询效率和可扩展性，使得知识图谱更易于应用和应用。In the above steps, Neo4j is selected as a graph database for storing feature knowledge graphs. The feature names and associated statistical indicators in different data texts are identified through NLP technology, and the association relationship between features is established, which is converted into nodes and edges in the graph to construct the network structure of the feature knowledge graph. In the present invention, a graph database, NLP entity recognition and relationship extraction technology are used to construct a feature knowledge graph, and the association relationship between features is presented in the form of a graph, which helps to better understand the relationship between features and provide deeper insights and support for data analysis, feature engineering and the establishment of machine learning models. At the same time, by establishing a graph index, the query efficiency and scalability of the graph can be improved, making the knowledge graph easier to apply and apply.
在本实施例中，应用图神经网络处理所述特征知识图谱，学习节点的独立特征及节点间的关联特征，包括以下步骤：In this embodiment, a graph neural network is applied to process the feature knowledge graph to learn the independent features of nodes and the associated features between nodes, including the following steps:
（1）基于特征知识图谱构造节点特征和边特征的特征向量，并选择GAT作为图神经网络架构；(1) Construct feature vectors of node features and edge features based on the feature knowledge graph, and select GAT as the graph neural network architecture;
（2）将图谱数据划分为训练集、验证集和测试集，定义损失函数，选择Adam优化器通过训练集进行图神经网络模型训练，通过图神经网络模型学习节点的独立特征和节点间的关联特征，通过验证集和测试集对图神经网络模型进行验证与调优；(2) Divide the graph data into training set, validation set and test set, define the loss function, select the Adam optimizer to train the graph neural network model through the training set, learn the independent features of nodes and the correlation features between nodes through the graph neural network model, and verify and tune the graph neural network model through the validation set and test set;
（3）利用训练好的图神经网络模型对不动产抵押物的风险进行评估。(3) Use the trained graph neural network model to assess the risk of real estate mortgages.
其中，所述图神经网络模型学习节点的独立特征和节点间的关联特征时，通过GNN层传播节点特征，每个节点聚合邻居节点信息，使用聚合的邻居节点信息更新每个节点的特征，并进行池化操作。Among them, when the graph neural network model learns the independent features of nodes and the correlation features between nodes, the node features are propagated through the GNN layer, each node aggregates the neighbor node information, uses the aggregated neighbor node information to update the features of each node, and performs pooling operations.
数据收集和预处理时，收集包括不动产信息、市场趋势、经济指标等多种数据，将数据转化为特征知识图谱，节点代表不动产、市场趋势、经济指标等，边表示它们之间的关联。图神经网络模型设计时，基于特征知识图谱构建节点和边的特征向量，设计一个GAT的图神经网络架构，在图神经网络中，通过多层的GNN层传播节点特征，聚合邻居信息，更新节点特征，并进行池化操作。During data collection and preprocessing, we collect a variety of data including real estate information, market trends, economic indicators, etc., and transform the data into a feature knowledge graph, where nodes represent real estate, market trends, economic indicators, etc., and edges represent the relationship between them. When designing the graph neural network model, we construct feature vectors of nodes and edges based on the feature knowledge graph, and design a GAT graph neural network architecture. In the graph neural network, we propagate node features through multiple GNN layers, aggregate neighbor information, update node features, and perform pooling operations.
在本实施例中，利用特征知识图谱中提取的特征训练风险评估模型，并通过训练的风险评估模型预测不动产抵押物的未来价值时，使用图神经网络模型作为特征提取器，提取的特征作为风险评估模型的输入；定义一个风险评估的对数损失函数，使用训练集数据训练风险评估模型，利用训练好的风险评估模型计算不动产抵押物的风险值，所述不动产抵押物的风险值包括违约概率和损失预期值，并基于市场趋势和经济指标预测不动产抵押物的未来价值，并根据计算出的风险值对不动产抵押物进行排序，形成风险评级。In this embodiment, the features extracted from the feature knowledge graph are used to train the risk assessment model, and when the future value of the real estate mortgage is predicted by the trained risk assessment model, a graph neural network model is used as a feature extractor, and the extracted features are used as the input of the risk assessment model; a logarithmic loss function for risk assessment is defined, the risk assessment model is trained using training set data, and the risk value of the real estate mortgage is calculated using the trained risk assessment model, wherein the risk value of the real estate mortgage includes the probability of default and the expected value of loss, and the future value of the real estate mortgage is predicted based on market trends and economic indicators, and the real estate mortgage is sorted according to the calculated risk value to form a risk rating.
通过结合图神经网络和特征知识图谱，本发明可以实现更精准、高效的不动产风险评估。本发明的上述方法利用了图神经网络在处理图数据时的优势，能够有效地捕捉节点间复杂的关系和特征信息，结合知识图谱中的领域知识，使得风险评估模型更具可解释性和泛化能力。通过上述方法可以更准确地评估不动产抵押物的风险，为决策者提供更全面的信息支持。By combining graph neural networks and feature knowledge graphs, the present invention can achieve more accurate and efficient real estate risk assessment. The above method of the present invention utilizes the advantages of graph neural networks in processing graph data, which can effectively capture the complex relationships and feature information between nodes, and combines the domain knowledge in the knowledge graph to make the risk assessment model more interpretable and generalizable. The above method can more accurately assess the risk of real estate collateral and provide more comprehensive information support for decision makers.
参见图4所示，在步骤4）中计算不动产抵押物的风险值，包括以下步骤：As shown in FIG4 , the risk value of the real estate mortgage is calculated in step 4), including the following steps:
步骤1、特征提取：通过图神经网络从知识图谱中提取出每个不动产抵押物节点的特征向量；设fi表示第i个不动产抵押物的特征向量；Step 1, feature extraction: extract the feature vector of each real estate mortgage node from the knowledge graph through the graph neural network; letfi represent the feature vector of the i-th real estate mortgage;
步骤2、构建评估模型：使用逻辑回归分类模型对违约概率（DP）进行评估，使用线性回归模型对损失预期值（LGD）进行评估；Step 2: Construct an evaluation model: Use a logistic regression classification model to evaluate the probability of default (DP), and use a linear regression model to evaluate the loss expectancy (LGD);
步骤3、模型训练：使用训练集数据训练违约概率（DP）模型和损失预期值（LGD）模型，分别预测违约概率和损失预期值；Step 3: Model training: Use the training set data to train the default probability (DP) model and the loss expected value (LGD) model to predict the default probability and loss expected value respectively;
步骤4、计算风险值：通过违约概率和损失预期值的组合来计算风险值（RiskValue,RV）；Step 4: Calculate the risk value: Calculate the risk value (RiskValue, RV) by combining the probability of default and the expected value of loss;
步骤5、风险值排序：根据计算出的风险值，对所有不动产抵押物进行排序，风险值高的排在前面，风险值低的排在后面，形成风险评估模型的输出。Step 5: Risk value sorting: All real estate mortgages are sorted according to the calculated risk values, with those with high risk values at the front and those with low risk values at the back, forming the output of the risk assessment model.
其中，违约概率（DP）模型表示为：Among them, the default probability (DP) model is expressed as:
； ;
式中，σ为sigmoid函数，用于将线性输出映射到概率空间（0到1之间）；Where σ is the sigmoid function, which is used to map the linear output to the probability space (between 0 and 1);
W为违约概率模型权重；fi第i个不动产抵押物的特征向量；b为偏置项。W is the weight of the default probability model; fiis the characteristic vector of the i-th real estate mortgage; b is the bias term.
其中，损失预期值（LGD）模型表示为：Among them, the loss expected value (LGD) model is expressed as:
； ;
式中，W和b为通过最小化损失函数学习的模型参数。Where W and b are the model parameters learned by minimizing the loss function.
在本实施例中，所述通过违约概率和损失预期值的组合来计算风险值，计算公式如下：In this embodiment, the risk value is calculated by combining the default probability and the loss expectation value, and the calculation formula is as follows:
； ;
式中，RVi是第i个不动产抵押物的总体风险值。Where RVi is the overall risk value of the i-th real estate mortgage.
示例性的，假设有一个包含100个不同抵押物的不动产投资组合，每个抵押物都有其特定的属性，如地理位置、建筑面积、建造年份、历史交易价格、租赁状态、历史违约记录等。那么，在特征提取阶段，使用地理信息系统（GIS）和市场交易数据库建立起一个不动产抵押物的知识图谱。然后，利用图神经网络（GNN）提取每个抵押物的特征向量fi。比如，对于第一个抵押物（记为不动产A），特征向量包含以下信息：位置热度评分、建筑面积1000平方米、建造年份2000年、历史交易价格150万美元、当前未出租、无历史违约记录。For example, suppose there is a real estate portfolio containing 100 different collaterals, each of which has its own specific attributes, such as geographic location, building area, construction year, historical transaction price, rental status, historical default record, etc. Then, in the feature extraction stage, a knowledge graph of real estate collateral is established using the Geographic Information System (GIS) and the market transaction database. Then, the feature vectorfi of each collateral is extracted using the Graph Neural Network (GNN). For example, for the first collateral (denoted as real estate A), the feature vector contains the following information: location heat score, building area of 1,000 square meters, construction year of 2000, historical transaction price of 1.5 million US dollars, currently not rented, and no historical default record.
然后，使用历史数据训练两个模型：违约概率（DP）模型和损失预期值（LGD）模型。其中，违约概率（DP）模型采用逻辑回归模型，输入不动产A的特征向量fA，训练出违约概率模型权重W以及偏置项b的参数。损失预期值（LGD）模型采用线性回归模型，输入同样的特征向量fA，训练出损失预期值模型权重W以及偏置项b的参数。通过收集过去10年内所有抵押物的违约和损失数据，以此为训练集来训练上述的DP模型和LGD模型。Then, two models are trained using historical data: the probability of default (DP) model and the loss expectancy (LGD) model. The probability of default (DP) model uses a logistic regression model, inputs the feature vector fA of real estate A, and trains the weights W of the probability of default model and the parameters of the bias term b. The loss expectancy (LGD) model uses a linear regression model, inputs the same feature vector fA , and trains the weights W of the loss expectancy model and the parameters of the bias term b. The default and loss data of all collaterals in the past 10 years are collected and used as training sets to train the above DP model and LGD model.
然后对不动产A计算风险值。首先计算违约概率DPA，然后计算损失预期值LGDA。假设DPA=0.03（即3%的违约概率），LGDA=0.5（即违约时损失为抵押物价值的一半），则不动产A的风险值RVA为0.03×0.5=0.015。Then calculate the risk value for real estate A. First calculate the default probability DPA , and then calculate the loss expectation valueLGDA . Assuming DPA = 0.03 (i.e. 3% default probability),LGDA = 0.5 (i.e. the loss in case of default is half of the value of the collateral), the risk value RVA of real estate A is 0.03 × 0.5 = 0.015.
最后进行风险值排序，对投资组合中的所有100个不动产抵押物进行风险值计算，并按照风险值从高到低进行排序。发现不动产A在整个投资组合中的风险值排名是第30位。最后根据这些计算结果，投资组合的管理者可以做出更明智的决策，比如对高风险的不动产实施更严格的借款条件，或者在市场波动时优先出售风险较低的资产。对于风险值较高的抵押物，可能还需要进一步分析其背后的原因，并采取相应的风险缓解措施，如增加抵押品价值、改善物业管理，或者通过适当的金融工具进行风险对冲。Finally, the risk value is sorted, and the risk value of all 100 real estate collaterals in the portfolio are calculated and sorted from high to low according to the risk value. It is found that the risk value of real estate A ranks 30th in the entire portfolio. Finally, based on these calculation results, portfolio managers can make more informed decisions, such as imposing stricter borrowing conditions on high-risk real estate, or giving priority to selling lower-risk assets when the market fluctuates. For collaterals with higher risk values, it may be necessary to further analyze the reasons behind them and take corresponding risk mitigation measures, such as increasing the value of collateral, improving property management, or hedging risks through appropriate financial instruments.
通过以上的实例可以看出，本发明的多维动态评估的不动产抵押物金融风险防控方法提供了一个全面评估和管理不动产抵押物金融风险的方案，有助于金融机构在复杂多变的市场环境中做出更为科学和合理的决策。It can be seen from the above examples that the multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method of the present invention provides a solution for comprehensive assessment and management of real estate mortgage financial risks, which helps financial institutions make more scientific and reasonable decisions in a complex and changing market environment.
综上所述，本发明给出了一种多维动态评估的不动产抵押物金融风险防控方法，结合了图神经网络、逻辑回归分类模型和线性回归模型，以更全面、准确地评估不动产抵押物的金融风险。利用图神经网络从知识图谱中提取每个不动产抵押物节点的特征向量，包括但不限于房屋面积、地理位置、建造年份等特征。这些特征向量为后续风险评估提供了基础数据。使用逻辑回归分类模型评估每个不动产抵押物的违约概率（DP），以及使用线性回归模型评估损失预期值（LGD）。逻辑回归模型通过特征向量预测不动产抵押物是否有违约风险，而线性回归模型则评估在发生违约时的损失程度。In summary, the present invention provides a multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method, which combines graph neural network, logistic regression classification model and linear regression model to more comprehensively and accurately assess the financial risk of real estate mortgage. The graph neural network is used to extract the feature vector of each real estate mortgage node from the knowledge graph, including but not limited to features such as house area, geographical location, and year of construction. These feature vectors provide basic data for subsequent risk assessment. The logistic regression classification model is used to assess the probability of default (DP) of each real estate mortgage, and the linear regression model is used to assess the loss expectation (LGD). The logistic regression model predicts whether the real estate mortgage has a risk of default through the feature vector, while the linear regression model assesses the degree of loss in the event of default.
利用训练集数据对违约概率（DP）模型和损失预期值（LGD）模型进行训练，以使其能够准确预测不同不动产抵押物的违约概率和损失预期值。通过结合违约概率和损失预期值的组合，根据给定的风险值计算公式，计算每个不动产抵押物的总体风险值。根据计算出的风险值，对所有不动产抵押物进行排序，将风险值高的排在前面，风险值低的排在后面，形成风险评估模型的输出。这样的排序可以帮助金融机构或相关机构更好地了解其不动产抵押物组合的风险分布情况，有针对性地采取风险管理措施。The default probability (DP) model and the loss expectation value (LGD) model are trained using the training set data to enable them to accurately predict the default probability and loss expectation value of different real estate collaterals. By combining the combination of default probability and loss expectation value, the overall risk value of each real estate collateral is calculated according to the given risk value calculation formula. According to the calculated risk value, all real estate collaterals are sorted, with high risk values at the front and low risk values at the back, forming the output of the risk assessment model. Such sorting can help financial institutions or related institutions better understand the risk distribution of their real estate collateral portfolios and take targeted risk management measures.
通过本发明的多维动态评估的不动产抵押物金融风险防控方法，金融机构可以更准确地评估不动产抵押物的风险，提前识别潜在的违约风险，并采取相应的风险防控措施，以降低金融风险并保护其利益。这种方法结合了多种技术手段，从特征提取到风险值计算，形成了一个完整的风险评估框架，为金融行业提供了一种全面、高效的不动产抵押物风险管理解决方案。Through the multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method of the present invention, financial institutions can more accurately assess the risks of real estate mortgages, identify potential default risks in advance, and take corresponding risk prevention and control measures to reduce financial risks and protect their interests. This method combines a variety of technical means, from feature extraction to risk value calculation, to form a complete risk assessment framework, providing a comprehensive and efficient real estate mortgage risk management solution for the financial industry.
应该理解的是，上述虽然是按照某一顺序描述的，但是这些步骤并不是必然按照上述顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，本实施例的一部分步骤可以包括多个步骤或者多个阶段，这些步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤中的步骤或者阶段的至少一部分轮流或者交替地执行。It should be understood that, although described in a certain order, these steps are not necessarily performed in sequence in the above order. Unless there is clear explanation in this article, the execution of these steps does not have strict order restriction, and these steps can be performed in other orders. Moreover, a part of the steps of the present embodiment may include a plurality of steps or a plurality of stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a part of the steps or stages in other steps or other steps.
在实施例中，本发明还提供了一种多维动态评估的不动产抵押物金融风险防控系统，旨在利用先进的数据分析和机器学习技术，对抵押物进行准确的风险评估和有效的风险预警防控，通过集成的模块化设计，将多维度数据转化为可操作的风险评估信息，并提供预警机制来降低潜在风险，该系统包括以下组件：In an embodiment, the present invention also provides a multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control system, which aims to use advanced data analysis and machine learning technology to accurately assess the risk of mortgages and effectively prevent and control risks. Through an integrated modular design, multi-dimensional data is converted into actionable risk assessment information, and an early warning mechanism is provided to reduce potential risks. The system includes the following components:
数据收集模块，用于收集与不动产抵押物贷款相关的多维数据，并进行数据清洗和预处理。The data collection module is used to collect multi-dimensional data related to real estate mortgage loans and perform data cleaning and pre-processing.
特征工程模块，用于根据预处理后的多维数据，确定不动产抵押物的关键特征要素，并作为知识图谱中的节点。The feature engineering module is used to determine the key characteristic elements of real estate mortgages based on the preprocessed multi-dimensional data and use them as nodes in the knowledge graph.
知识图谱构建模块，用于运用实体识别和关系抽取建立所述节点之间的关联关系，构建特征知识图谱。The knowledge graph construction module is used to use entity recognition and relationship extraction to establish the association relationship between the nodes and construct a feature knowledge graph.
图神经网络处理模块，用于应用图神经网络处理所述特征知识图谱，学习节点的独立特征及节点间的关联特征。The graph neural network processing module is used to apply the graph neural network to process the feature knowledge graph and learn the independent features of the nodes and the correlation features between the nodes.
风险评估模块，用于利用特征知识图谱中提取的特征训练风险评估模型，并通过训练的风险评估模型预测不动产抵押物的未来价值，基于节点特征计算风险值，对风险值进行排序，形成风险评估模型的输出。The risk assessment module is used to train the risk assessment model using the features extracted from the feature knowledge graph, and predict the future value of the real estate mortgage through the trained risk assessment model, calculate the risk value based on the node features, sort the risk values, and form the output of the risk assessment model.
风险预警防控模块，用于将风险评估模型的输出与事先设定的风险阈值进行比较，一旦监测到任何风险值超过风险阈值时，自动触发预警机制，生成风险评估结果并启动风险防控措施，同时根据风险评估结果调整金融机构的抵押贷款政策。The risk early warning and control module is used to compare the output of the risk assessment model with the pre-set risk threshold. Once any risk value is detected to exceed the risk threshold, the early warning mechanism is automatically triggered to generate risk assessment results and initiate risk prevention and control measures. At the same time, the mortgage loan policy of the financial institution is adjusted according to the risk assessment results.
在本实施例中，所述特征工程模块中包括：In this embodiment, the feature engineering module includes:
统计分析单元，用于运行describe（）等函数进行数据总结统计；Statistical analysis unit, used to run describe() and other functions to summarize data;
相关性分析单元，用于计算皮尔逊相关系数并生成热图；Correlation analysis unit, used to calculate Pearson correlation coefficient and generate heat map;
单变量分析单元，用于通过ANOVA测试筛选显著特征；Univariate analysis unit, used to screen significant features through ANOVA test;
特征选择单元，用于识别关键特征；A feature selection unit, used to identify key features;
交叉验证单元，用于通过k-fold交叉验证特征的稳定性和预测能力。The cross-validation unit is used to verify the stability and predictive power of features through k-fold cross-validation.
在本实施例中，所述知识图谱构建模块包括：In this embodiment, the knowledge graph construction module includes:
图数据库，用于存储和管理知识图谱数据。Graph database, used to store and manage knowledge graph data.
NLP模型，用于实体识别和关系抽取。NLP models for entity recognition and relation extraction.
实体关联对齐单元，用于合并不同来源的同一实体表示。Entity association alignment unit, used to merge the same entity representations from different sources.
知识图谱索引构建单元，用于提高查询效率。Knowledge graph index building unit, used to improve query efficiency.
在本实施例中，所述风险评估模块中采用图神经网络作为特征提取器，用于从知识图谱提取风险特征；风险评估的对数损失函数定义，用于指导模型训练；违约概率模型和损失预期值模型用于风险值评估。In this embodiment, the risk assessment module uses a graph neural network as a feature extractor to extract risk features from the knowledge graph; the logarithmic loss function of risk assessment is defined to guide model training; the default probability model and the loss expectation value model are used for risk value assessment.
在本实施例中，所述风险预警防控模块中包括：In this embodiment, the risk warning and control module includes:
风险阈值设定单元，用于设定风险预警阈值；A risk threshold setting unit, used to set a risk warning threshold;
风险评估结果处理单元，用于风险值排序和输出；以及A risk assessment result processing unit for ranking and outputting risk values; and
预警机制触发器，用于在风险值超过阈值时发出预警并采取相应措施。The early warning mechanism trigger is used to issue an early warning and take corresponding measures when the risk value exceeds the threshold.
在本实施例中，所述多维动态评估的不动产抵押物金融风险防控系统，还包括策略调整模块，用于根据风险评估结果调整金融机构策略。In this embodiment, the multi-dimensional dynamic assessment system for preventing and controlling financial risks of real estate mortgages further includes a strategy adjustment module for adjusting the strategy of financial institutions according to the risk assessment results.
在本实施例中，多维动态评估的不动产抵押物金融风险防控系统在执行时采用如前述的一种多维动态评估的不动产抵押物金融风险防控方法的步骤，因此，本实施例中对多维动态评估的不动产抵押物金融风险防控系统的运行过程不再详细介绍。In this embodiment, the multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control system adopts the steps of the aforementioned multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method during execution. Therefore, the operation process of the multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control system will not be introduced in detail in this embodiment.
本发明的多维动态评估的不动产抵押物金融风险防控系统，通过高度集成的模块化设计，允许金融机构以一种结构化和自动化的方式对不动产抵押物进行风险评估和管理，不仅提供了一种基于数据驱动的风险评估方法，而且还提供了实时的风险预警和防控机制，从而提高了金融机构在不动产抵押贷款业务中的风险管理能力。The multi-dimensional dynamic assessment of the real estate mortgage financial risk prevention and control system of the present invention, through a highly integrated modular design, allows financial institutions to conduct risk assessment and management of real estate mortgages in a structured and automated manner, and not only provides a data-driven risk assessment method, but also provides a real-time risk warning and prevention and control mechanism, thereby improving the risk management capabilities of financial institutions in real estate mortgage loan business.
在的实施例中，在本发明实施例的还提供了一种计算机设备，包括至少一个处理器，以及与所述至少一个处理器通信连接的存储器，所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器执行所述的多维动态评估的不动产抵押物金融风险防控方法，该处理器执行指令时实现上述多维动态评估的不动产抵押物金融风险防控方法实施例中的步骤。In an embodiment, a computer device is also provided in an embodiment of the present invention, including at least one processor and a memory communicatively connected to the at least one processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the at least one processor so that the at least one processor executes the multi-dimensional dynamic assessment of the real estate mortgage financial risk prevention and control method, and the processor implements the steps in the above-mentioned multi-dimensional dynamic assessment of the real estate mortgage financial risk prevention and control method embodiment when executing the instructions.
在的实施例中，提供了一种计算机可读存储介质，计算机可读存储介质存储有计算机指令，所述计算机指令用于使所述计算机执行所述的多维动态评估的不动产抵押物金融风险防控方法的步骤。In an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the steps of the multi-dimensional dynamic assessment of the real estate mortgage financial risk prevention and control method.
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机指令表征的计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和易失性存储器中的至少一种。A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be realized by instructing the relevant hardware through a computer program represented by computer instructions, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory.
非易失性存储器可包括只读存储器、磁带、软盘、闪存或光存储器等。易失性存储器可包括随机存取存储器或外部高速缓冲存储器。作为说明而非局限，RAM可以是多种形式，比如静态随机存取存储器或动态随机存取存储器等。Non-volatile memory may include read-only memory, magnetic tape, floppy disk, flash memory or optical storage, etc. Volatile memory may include random access memory or external cache memory. As an illustration and not limitation, RAM may be in various forms, such as static random access memory or dynamic random access memory, etc.
以上所述仅为本发明的较佳实施例而已，并不用以限制本发明，凡在本发明的精神和原则之内所做的任何修改、等同替换和改进等，均应包含在本发明的保护范围之内。The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410337984.6ACN117934162A (en) | 2024-03-25 | 2024-03-25 | Multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method and system |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410337984.6ACN117934162A (en) | 2024-03-25 | 2024-03-25 | Multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method and system |
| Publication Number | Publication Date |
|---|---|
| CN117934162Atrue CN117934162A (en) | 2024-04-26 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410337984.6APendingCN117934162A (en) | 2024-03-25 | 2024-03-25 | Multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method and system |
| Country | Link |
|---|---|
| CN (1) | CN117934162A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118628235A (en)* | 2024-08-14 | 2024-09-10 | 无锡锡商银行股份有限公司 | A risk assessment method for housing mortgage loans |
| CN119647947A (en)* | 2024-11-25 | 2025-03-18 | 国网江苏省电力有限公司建设分公司 | A method and system for engineering supervision data management based on deep learning |
| CN119762218A (en)* | 2024-12-27 | 2025-04-04 | 中国工商银行股份有限公司 | Risk prediction method, device, storage medium and electronic device for loan business |
| CN120181330A (en)* | 2025-05-19 | 2025-06-20 | 中国计量大学 | Real estate seizure probability prediction method, device and readable storage medium based on multi-time span feature integration and explainable machine learning algorithm |
| CN120338945A (en)* | 2025-04-02 | 2025-07-18 | 国耀融汇融资租赁有限公司 | A method and system for predicting financial leasing risks based on knowledge graph |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118628235A (en)* | 2024-08-14 | 2024-09-10 | 无锡锡商银行股份有限公司 | A risk assessment method for housing mortgage loans |
| CN119647947A (en)* | 2024-11-25 | 2025-03-18 | 国网江苏省电力有限公司建设分公司 | A method and system for engineering supervision data management based on deep learning |
| CN119647947B (en)* | 2024-11-25 | 2025-09-16 | 国网江苏省电力有限公司建设分公司 | A method and system for engineering supervision data management based on deep learning |
| CN119762218A (en)* | 2024-12-27 | 2025-04-04 | 中国工商银行股份有限公司 | Risk prediction method, device, storage medium and electronic device for loan business |
| CN120338945A (en)* | 2025-04-02 | 2025-07-18 | 国耀融汇融资租赁有限公司 | A method and system for predicting financial leasing risks based on knowledge graph |
| CN120181330A (en)* | 2025-05-19 | 2025-06-20 | 中国计量大学 | Real estate seizure probability prediction method, device and readable storage medium based on multi-time span feature integration and explainable machine learning algorithm |
| Publication | Publication Date | Title |
|---|---|---|
| Van Thiel et al. | Artificial intelligence credit risk prediction: An empirical study of analytical artificial intelligence tools for credit risk prediction in a digital era | |
| Cho et al. | A hybrid approach based on the combination of variable selection using decision trees and case-based reasoning using the Mahalanobis distance: For bankruptcy prediction | |
| CN117934162A (en) | Multi-dimensional dynamic assessment of real estate mortgage financial risk prevention and control method and system | |
| Van Thiel et al. | Artificial Intelligent Credit Risk Prediction: An Empirical Study of Analytical Artificial Intelligence Tools for Credit Risk Prediction in a Digital Era. | |
| KR102499182B1 (en) | Loan regular auditing system using artificia intellicence | |
| CN119444397A (en) | Credit risk assessment model construction method and credit risk assessment model | |
| CN114612239A (en) | Stock public opinion monitoring and wind control system based on algorithm, big data and artificial intelligence | |
| CN118037304A (en) | A financial risk level labeling method and system based on data mining | |
| KR102596740B1 (en) | Method for predicting macroeconomic factors and stock returns in the context of economic uncertainty news sentiment using machine learning | |
| CN118734207A (en) | Market economic risk data processing system and method based on artificial intelligence | |
| CN119379428A (en) | A method for dynamic monitoring and report generation of financial risks based on multi-source data | |
| CN112766814A (en) | Training method, device and equipment for credit risk pressure test model | |
| Lin et al. | Research on credit big data algorithm based on logistic regression | |
| du Toit et al. | Shapley values as an interpretability technique in credit scoring | |
| CN113393316A (en) | Loan overall process accurate wind control and management system based on massive big data and core algorithm | |
| Zang | Construction of Mobile Internet Financial Risk Cautioning Framework Based on BP Neural Network | |
| CN117172910A (en) | Credit evaluation method and device based on EBM model, electronic equipment and storage medium | |
| CN117764692A (en) | Method for predicting credit risk default probability | |
| Liu et al. | Investment decision support for engineering projects based on risk correlation analysis | |
| Ouyang | Financial Risk Control of Listed Enterprises Based on Risk Warning Model | |
| Hu | Development of a Machine Learning-Based Financial Risk Control System | |
| CN120317882B (en) | Method and system for constructing social credit index library based on big data | |
| CN119962996B (en) | Asset screening and management method and system based on automated rules | |
| Lakshmi et al. | Machine learning approach for taxation analysis using classification techniques | |
| Chen et al. | Construction of Bank Credit White List Access System Based on Grey Clustering Algorithm |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20240426 | |
| RJ01 | Rejection of invention patent application after publication |