Detailed Description
The foregoing description is only an overview of the present application, and is intended to be implemented in accordance with the teachings of the present application in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present application more readily apparent.
The present application will be described in further detail below with reference to the accompanying drawings, in order to make the objects, technical solutions and advantages of the present application more apparent, and the described embodiments should not be construed as limiting the present application, but all other embodiments obtained by those skilled in the art without making any inventive effort are within the scope of the present application.
In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict, the term "first\second" being referred to merely as distinguishing between similar objects and not representing a particular ordering for the objects. The terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server that comprises a list of steps or elements is not necessarily limited to those steps or elements that are expressly listed or inherent to such process, method, article, or apparatus, but may include other steps or modules that may not be expressly listed or inherent to such process, method, article, or apparatus, and unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used herein is for the purpose of describing embodiments of the application only.
The embodiment of the application provides a city governance scheduling method based on dynamic evaluation, as shown in fig. 1, comprising the following steps:
Step S100, a first spatial region is acquired, where the first spatial region refers to any region in a spatial region set obtained by dividing a target city, and the first spatial region has an identifier of a first functional characteristic. Specifically, firstly, the basis for dividing the target city is determined, and factors such as geographic characteristics, administrative regions, functional requirements and the like can be comprehensively considered. The division operation is implemented by using tools such as geographic information system technology and city planning software, for example, different regions are defined according to natural boundaries, administrative boundaries or functional requirements. After division, a space region set is obtained, one region is selected as a first space region, and a first functional characteristic identification is given according to the specific function and characteristics of the region, for example, a commercial region can be identified as frequent commercial activities, large traffic volume and the like, a residential region can be identified as concentrated residence, perfect living service facilities and the like, and an industrial region can be identified as concentrated factory enterprises, large cargo transportation and the like.
And step 200, analyzing the collected first region characteristic information of the first space region to obtain a first governance fitness of the first space region, which is determined to be the first functional characteristic. Specifically, the type of characteristic information of the first spatial region is first determined, including topography, traffic conditions, demographics, building facilities, public service facilities, and the like. Information is collected by means of geographic information systems, traffic monitoring equipment, census data, and the like. Then, an evaluation index system is established according to the first functional characteristic requirement, for example, the commercial district considers traffic convenience, population density, consumption capability, commercial facility matching degree and the like, and the residential district considers environmental quality, education medical resource richness, leisure and entertainment facility matching and the like. And according to the characteristic indexes, weighting and summing the scores of all indexes by using a method combining quantitative analysis and qualitative analysis to obtain the final total score of the treatment adaptability. The fitness score may be a numerical score (e.g., 1-100 points) or may be presented using a hierarchical classification (e.g., low, medium, high). And (3) analyzing the obtained first treatment fitness and helping to determine the fitness level.
In one possible implementation manner, the collected first region feature information of the first spatial region is analyzed to obtain a first governance fitness of the first spatial region determined to be the first functional characteristic, and step S200 further includes step S210, where a predetermined topography index is read. Specifically, the content of the predetermined topography index is determined, the predetermined topography index comprises specific elements such as altitude, basin, mountain peak and the like, the index can reflect natural geographic characteristics of a first space region, for example, the altitude can influence climate, environment and building design, the basin topography can influence air circulation and drainage systems, the mountain peak can limit traffic routes and landscape planning, a satellite image, a topography map and field investigation are combined, the satellite image can provide macroscopic topography features, the topography map can display detailed information such as contour lines and the like, and the field investigation can verify and supplement information of other data sources.
And step S220, carrying out multidimensional feature collection on the first space region based on the preset topography index to obtain first topography feature information. Specifically, the influence of the elevation on the first space area is analyzed, the climate difference of different elevation areas is considered, the higher the elevation is, the lower the air temperature is, the lower the air pressure is, for example, more heating and warming measures are needed in a high elevation area, the requirement of the elevation on the building design is evaluated, the influence of the elevation on the environment is analyzed, the vegetation type and the animal distribution of different altitudes are different, the influence of basin topography on the first space area is analyzed, the influence of basin topography on the air circulation is considered, the limitation of basin topography on the traffic plan is analyzed, mountains around the basin influence the layout and the construction cost of traffic lines, the influence of mountain peaks on the traffic lines is evaluated, the mountain peaks can become barriers to traffic, engineering facilities such as tunnels, bridges are needed to be constructed to overcome, and the analysis results of topography indicators such as the elevation, basin, mountain peaks are summarized and arranged through tools such as a Geographic Information System (GIS) and a remote sensing technology, and the first topography feature information is formed, including description, influence assessment and identification of different topography features and potential problems can be combined with historical data, satellites, ground topography and land topography feature analysis means.
Step S230, reading a predetermined flow index. Specifically, the content of the preset flow index is determined, the preset flow index comprises factors such as people, vehicle flow, distance, road conditions, traffic convenience level and the like, the index can reflect the people and vehicle flow conditions of a first space region, for example, the population density and activity intensity of a people flow quantity reflecting region, the vehicle flow quantity can reflect traffic jam conditions and road bearing capacity, the distance can measure the accessibility between different places, the road conditions comprise factors such as road width and road surface quality, the traffic convenience level considers factors such as public traffic coverage and traffic signal lamp setting, and the like, and the traffic monitoring equipment is used for acquiring vehicle flow quantity data.
And step S240, carrying out multidimensional feature collection on the first space region based on the preset flow index to obtain first flow feature information. The method comprises the steps of firstly evaluating influences of personnel and vehicle flows in different areas on the areas by analyzing time distribution, space distribution and flow quantity of the personnel and vehicle flows, and then evaluating influences of the personnel and vehicle flows on the areas by considering road conditions, traffic convenience and coverage conditions of public traffic. And carrying out weighting treatment and normalization treatment on the analysis results, comprehensively evaluating the influence of each flow characteristic, and obtaining final first flow characteristic information. The information may be presented in the form of numerical, hierarchical or weighted scores. For example, indexes such as personnel flow, vehicle flow, road condition, traffic convenience and the like are given different weights, and are combined with preset scoring standards (such as scores of 1-10 or low, medium and high grades) for weighting and summarizing, so that the flow characteristic score of the area is finally obtained. The comprehensive analysis will provide a comprehensive assessment of personnel flow, vehicle flow, road conditions, etc.
Step S250, analyzing the first topography feature information and the first flow feature information to obtain the first treatment fitness. Specifically, the influence of the topography and the flow characteristics is analyzed, the first topography characteristic information and the first flow characteristic information are compared and analyzed, and the interrelation and the influence between the first topography characteristic information and the first flow characteristic information are found out. In order to realize quantitative evaluation of the treatment fitness, a treatment fitness evaluation model can be established based on the analyzed first topography feature information and the first flow feature information, the model can be established by an Analytic Hierarchy Process (AHP) or a fuzzy comprehensive evaluation method, and the weight and evaluation standard of each factor are determined by combining multidimensional analysis factors such as traffic convenience, environmental quality, resource utilization efficiency and the like, and firstly, respective evaluation dimensions are respectively established according to the first topography feature information and the first flow feature information, and different weights are given. Then, the relative importance of each factor is calculated by comparing the influence degrees of different indexes by using an analytic hierarchy process, and the weight coefficient is determined. For the fuzzy comprehensive evaluation method, the fuzzy processing can be carried out on the evaluation dimension by constructing a fuzzy set and a membership function, the comprehensive evaluation is carried out on the treatment fitness of the first space region by combining expert judgment and actual data, the comprehensive evaluation of the first space region is carried out by inputting the evaluation results of the topography, the flow characteristics and other economic factors (such as traffic convenience degree, environmental resource endowment and the like) into a treatment fitness evaluation model, weighting and summing are carried out by using the weight coefficients of each dimension, and the overall fitness score is obtained by calculation.
In a possible implementation manner, the first topography feature information and the first flow feature information are analyzed to obtain the first governance fitness, step S250 further includes step S251 of calling up a municipal governance database, and extracting a first historical governance record in the municipal governance database, where the first historical governance record refers to a municipal planning scheduling record for scheduling a first historical space region governance as the first functional characteristic. Specifically, the function of the urban governance database is clarified, wherein governance records of different spatial regions in different periods are included, for example, the planning and scheduling process of converting a certain region from an industrial region to a commercial region is recorded in the database, the information including measures taken, problems encountered, final effects and the like is included, a first historical governance record is extracted, and urban governance records which schedule the first historical spatial region governance to have the same first functional characteristic as the current first spatial region, namely, first historical governance records, are screened from the urban governance database.
Step S252, constructing first historical topography feature information of the first historical space region based on the predetermined topography index. Specifically, the first historical space region is analyzed by using predetermined terrain and topography indexes determined previously, such as altitude, basin, mountain peak and the like, and the terrain and topography characteristics of the first historical space region are determined, for example, by referring to a historical map, geographical data, field investigation and the like, including whether the conditions of mountain, river, altitude and the like exist. The first historical topography and topography characteristic information specifically comprises 1) numerical information such as specific altitude, gradient of ground fluctuation, flow direction of a river, river basin range and the like in an area, 2) grade information such as whether the area belongs to a mountain area, a plain or a hilly area, is divided into 'simple' or 'complex' according to topography complexity, and the like, and 3) descriptive information such as whether special geographic phenomena (such as areas rich in water sources or areas vulnerable to flood) exist. And finally, sorting and summarizing the analyzed topographic and topographic feature information to form first historical topographic and topographic feature information.
Step S253, performing similarity analysis on the first topography feature information and the first historical topography feature information to obtain a first topography similarity. Specifically, feature information comparison is performed, first topographic feature information of the current first spatial region is compared with first historical topographic feature information of the first historical spatial region one by one, for example, altitude ranges of the first topographic feature information and the first topographic feature information are compared, whether features such as basin or mountain peak are present or not is compared, similarity calculation is performed, an appropriate similarity calculation method such as cosine similarity and euclidean distance is adopted, quantization calculation is performed on the two topographic feature information, and the first historical topographic feature information is obtained by quantizing topographic features of the current first spatial region and the first historical spatial region and comprises aspects such as altitude, topographic type and gradient. The elevation can be represented by the difference between the maximum elevation and the minimum elevation, the terrain type can be encoded into different values, the gradient is quantified by calculating the average gradient of the area, the quantified terrain feature data is converted into feature vectors, the first situational similarity, the terrain features of the current area and the historical area are compared by adopting a similarity calculation method such as cosine similarity or Euclidean distance, the first situational similarity is obtained by calculating the included angle or distance between the feature vectors of the current area and the historical area, the first situational similarity is obtained, the value reflects the similarity of the current area and the historical area on the terrain features, and the larger the value or the smaller the distance is, the higher the similarity of the current area and the historical area is represented.
Step S254, when the first situation similarity meets a predetermined similarity threshold, extracting a first historical governance difficulty in the first historical governance record. Specifically, the first historical governance difficulty refers to a comprehensive assessment of various difficulties and challenges encountered in implementing governance measures for a particular spatial region during past governance procedures. Such difficulties include, but are not limited to, 1) geographical and environmental factors such as complex terrain within an area, extreme weather, environmental protection requirements, etc., which may result in increased project costs or delays in progress of implementation, 2) social factors such as population changes, demographics, industry development, 3) economic and technical difficulties such as construction costs that are excessive, technical problems, infrastructure construction bottlenecks, etc. Each difficulty may be represented by a numerical scoring criteria, typically using a scale from 1 to 5 or a specific quantitative indicator (e.g., increased construction cost percentage, length of delay, etc.) to represent the severity of the difficulty, and finally a weighted sum is taken to obtain a composite value as the first historical governance difficulty.
And S255, taking the reciprocal of the first historical governance difficulty as the first governance situation coupling degree. The method comprises the steps of determining the relation between difficulty and coupling degree, wherein the difficulty and the coupling degree are in an inverse relation, namely, the higher the governance difficulty is, the lower the governance situation coupling degree is, otherwise, the lower the governance difficulty is, the higher the governance situation coupling degree is, for example, if the first historical governance difficulty is 0.6, the first governance situation coupling degree is 1/0.6 approximately equal to 1.67, the first governance situation coupling degree is determined, the reciprocal of the first historical governance difficulty is calculated, the first governance situation coupling degree is obtained, the coupling degree reflects the governance similarity and the borrowability of the current first space region and the first historical space region in terms of topography and topography, and for example, if the first governance situation coupling degree is higher, the current region is similar to the historical region in terms of topography and topography.
And step S256, performing mutation weighting treatment on the plurality of flow characteristic parameters in the first flow characteristic information to obtain a first governance flow coupling degree. Specifically, a plurality of key flow characteristic parameters, such as a person flow amount, a vehicle flow amount, a traffic distance, a traffic convenience degree, and the like, are selected from the first flow characteristic information. For example, the amount of person flowing is represented by the number of people passing through the area per day, and the amount of vehicle flowing is represented by the number of vehicles passing through per hour. And according to the importance of the flow characteristic parameters and the influence of the variation degree, giving different weights to each parameter. The weight is determined based on the relative importance of the flow characteristics to the current zone remediation goals, as well as the variance differences from the historical zones. For example, if the degree of traffic convenience has a greater impact on the area and there is a significant difference from the history area, the parameter may be given a higher weight. And (3) integrating the values of the flow characteristic parameters through weighted calculation to obtain a first governance flow coupling degree, namely, the integrated coupling degree between the flow characteristic and the governance effect.
Step S257, taking the average value of the first governance situation coupling degree and the first governance flow coupling degree as the first governance fitness. Specifically, the topography and the flow characteristics are comprehensively considered, the first governance situation coupling degree reflects governance similarity in the topography and the flow characteristics, the first governance flow coupling degree reflects governance similarity in the flow characteristics, the factors in the two aspects can be comprehensively considered by taking the average value of the first governance situation coupling degree and the flow characteristics to obtain the first governance adaptation degree, for example, if the first governance situation coupling degree is 1.67 and the first governance flow coupling degree is 1.2, the first governance adaptation degree is (1.67+1.2)/2=1.435, the first governance adaptation degree is determined, the first governance adaptation degree can be used for measuring governance feasibility and adaptation of the current first space region as the first functional characteristic, and the higher the adaptation degree is, the higher the governance adaptation degree is, the region is similar to the historical successful governance region in the topography and the flow characteristics, and the governance difficulty is relatively low.
And step S300, when the first treatment fitness meets a preset fitness threshold, a density prediction analysis is carried out on the first space region by calling a density database, so as to obtain a first prediction density. Specifically, a predetermined fitness threshold is determined according to a target and multiple factors, the calculated first treatment fitness is compared with the predetermined fitness threshold, if the calculated first treatment fitness is greater than or equal to the threshold, relevant data is called from a density database according to the identification and the functional characteristics of the first space region, the database comprises historical density records of a plurality of historical space regions consistent with the first functional characteristics and relevant influence factor records, and the data structure is organized according to time, region identification and the like. Then, a suitable prediction method, such as time series analysis, regression analysis or machine learning algorithm, is selected, and the data is preprocessed, including cleaning of missing and outliers, normalization and feature engineering. The selected method and the preprocessed data are then used to predict, e.g., future concentration by time series model input and historical data. Finally, a first predicted concentration, which includes a person predicted concentration, a vehicle predicted concentration, and the like, is obtained, and may be represented by a numerical value or a grade.
In one possible implementation manner, when the first governance fitness meets a predetermined fitness threshold, a density prediction analysis is performed on the first spatial region by using a density database, so as to obtain a first predicted density, step S300 further includes step S310 of extracting a first density time sequence in the density database, where the density database includes historical density record data of a plurality of historical spatial regions consistent with the first functional characteristic, and the first density time sequence includes a first vehicle density time sequence and a first personnel density time sequence. Specifically, the concentration database is a collection of historical concentration records of spatial regions related to different functional characteristics, and covers people and vehicle concentration data of a plurality of historical spatial regions at different time points. Data is critical to predicting the concentration of first spatial regions having the same functional characteristics. For example, for the business segment functional characteristics of a city, the database contains personnel and vehicle concentration data for a plurality of mature business segments over different time periods, including weekdays, holidays, and different seasonal variations, etc. And respectively recording the change conditions of the number of vehicles and the number of personnel in the relevant historical area along with time by extracting a first concentration time sequence corresponding to the historical concentration record data. The first vehicle concentration time sequence is collected by traffic monitoring equipment, parking lot records and the like, and the first personnel concentration time sequence is obtained by census data, traffic monitoring equipment, mobile equipment positioning data and the like. Based on the time series data, a predictive algorithm (e.g., a time series analysis or regression model) is used, with the first predictive concentration referring to predicting the person and vehicle flow densities for the area over a period of time in the future based on the historical concentration data.
In step S320, an intelligent prediction model is activated, the intelligent prediction model including a vehicle prediction layer and a person prediction layer. Specifically, an intelligent prediction model is constructed, data related to traffic flow data, geometric characteristics of roads, land utilization types of surrounding areas and weather conditions of different time periods are collected, the collected data comprises demographic information, building function information, time characteristics, public transportation line and station information and activity event arrangement in the areas, the collected data is sorted and cleaned, abnormal values and missing values are removed, quality and integrity of the data are ensured, characteristic engineering processing is carried out on the sorted data, useful characteristics are extracted to improve the prediction capability of the model, time characteristics are constructed in terms of vehicle concentration, such as dividing one day into different time periods, one week into working days and weekends, time information is quantitatively coded, so that the model can learn the influence of different time periods on the vehicle concentration, and space characteristics are extracted, such as calculating congestion index, traffic information and traffic information, Road connectivity, etc., weather conditions are classified and encoded, such as sunny days, cloudy days, rainy days, snowy days, etc., respectively represented by different values, population feature vectors are constructed in terms of personnel concentration, statistics including mean values of age distribution, standard deviation, etc., proportion of professional structures, etc., business activity indexes, educational resource concentration, etc., are calculated according to building function information, activity events are quantized, such as setting different grades according to activity scale and predicted participation number, an appropriate model algorithm is selected according to data characteristics and requirements of prediction tasks, and for vehicle concentration prediction, a time sequence analysis algorithm such as ARIMA (autoregressive integral sliding average model) is used to capture trends in time sequence data, For people concentration prediction, a classification and regression algorithm based on machine learning is used, for example, a decision tree algorithm branches according to different characteristic conditions, a prediction model of people concentration is constructed, better interpretability is achieved, and a sorted data set is divided into a training set, a training set and a training set, the method comprises the steps of verifying a set and a test set, wherein the training set is used for learning and parameter adjustment of a model, the verifying set is used for evaluating performance of the model in a training process and selecting optimal super parameters, the test set is used for finally evaluating generalization capability of the model, vehicle concentration prediction is taken as an example, an LSTM model is used, parameters of the model are initialized first, first vehicle multi-domain characteristic information in the training set is sequentially input into the LSTM model according to time, the model continuously adjusts the parameters through a back propagation algorithm to minimize errors between predicted vehicle concentration and actual vehicle concentration, for personnel concentration prediction, similarly, the first person multi-domain characteristic information in the training set is input into the selected model for training, for example, in decision tree training, the model is continuously split according to different values of the characteristics, a tree structure is constructed, each leaf node corresponds to the same personnel concentration level or similar personnel quantity range as much as possible, the prediction error of the model on the training set is minimized through continuously adjusting the split conditions and parameters, and the verification and the optimal testing are carried out on the training set, and the final evaluation is carried out on the training set.
Step S330, analyzing, by the vehicle prediction layer, the first vehicle multi-domain feature information of the first vehicle concentration sequence, to obtain a first predicted vehicle concentration. Specifically, the first vehicle multi-domain feature information is comprehensive data for describing vehicle concentration related features, and is obtained through the previous feature engineering processing, including time features, space features, traffic related features and weather features, the vehicle prediction layer analyzes the first vehicle multi-domain feature information by using a trained vehicle concentration prediction model, for example, if an LSTM model is used, the first vehicle multi-domain feature information at the current moment is input into the model, the model calculates a vehicle concentration predicted value in a future time period according to a previously learned time sequence mode and a feature relation, namely, the first predicted vehicle concentration, the LSTM model can effectively process a long-term dependency relation in the time sequence data through a memory unit and a gating mechanism in the LSTM model, accurately captures a change trend of the vehicle concentration along with time and influence of various factors on the first vehicle multi-domain feature information, and in the prediction process, the model gradually updates an internal state according to the input feature information and finally outputs a prediction result.
And step S340, analyzing the first person multi-domain characteristic information of the first person concentration time sequence through the person prediction layer to obtain a first predicted person concentration. Specifically, the first personnel multi-domain feature information is data related to personnel concentration after feature engineering processing, including time features, space features, demographic features, public transportation features and activity event features, the personnel prediction layer processes the first personnel multi-domain feature information according to a trained personnel concentration prediction model, a random forest model is taken as an example, the first personnel multi-domain feature information is input into a plurality of decision trees for parallel prediction, each decision tree branches and judges according to different combinations of features, a personnel concentration prediction result is finally obtained, then the prediction results of the decision trees are integrated to obtain the final first prediction personnel concentration, and the random forest model can reduce the risk of overfitting of a single decision tree and improve the accuracy and stability of prediction by integrating the prediction capabilities of the decision trees.
Step S350, the first predicted vehicle concentration and the first predicted person concentration together form the first predicted concentration. Specifically, the first predicted vehicle concentration obtained by the vehicle prediction layer and the first predicted personnel concentration obtained by the personnel prediction layer are combined, the two prediction results reflect the concentration condition of the first space region from two aspects of the vehicle and the personnel respectively, so as to jointly form the first prediction concentration, and the two prediction results can be simply combined into a vector form, for example [ the first prediction vehicle concentration, the first prediction personnel concentration ], or a comprehensive concentration index is calculated according to actual demands, for example, the number of the vehicles and the personnel is weighted and summed to obtain a single numerical value to represent the first prediction concentration, the first prediction concentration can be represented and applied in various modes, depending on the actual demands, and in the aspect of traffic management, if the first prediction concentration of the first space region is predicted to be higher in a certain time period, the traffic management department can formulate a traffic guiding scheme in advance, such as adjusting the time allocation of traffic lights, increasing the number of traffic lights, guiding vehicles to split, and the like, and simultaneously, the traffic lights can be reasonably planned to advance the parking space of the vehicles to reserve the parking lot according to the first prediction concentration, or the parking lot can be guided to the parking lot in advance.
In one possible implementation, an intelligent predictive model is activated, the intelligent predictive model including a vehicle predictive layer and a person predictive layer, and step S320 further includes step S321, reading a predetermined unit time zone. Specifically, a predetermined unit time zone is determined, the selection of the predetermined unit time zone needs to be determined according to specific study objects and prediction targets, different unit time zones are suitable for different scenes and analysis requirements, if weekly periodic changes of the area, such as the difference between weekends and working days, are studied, the unit time zone is more suitable, the characteristic of concentration of different dates in the week is observed, and for some long-term trend analysis, such as the unit time zone of months or quarters, the seasonal or periodic law is more helpful to be found.
And S322, slicing the first concentration time sequence by taking the preset unit time zone as a constraint to obtain a first time sequence slicing set. Specifically, the first time series is a time-varying record sequence of people and vehicles in the first space region, and includes time-ordered time series of time series data, which are recorded at smaller time intervals such as hours, minutes, etc., for example, in the case of a month of first time series data, each hour has a corresponding people and vehicle time series record, a long time series of time series data is formed, a predetermined unit time zone is used as a constraint, the first time series is divided into a plurality of segments, if the predetermined unit time zone is a day, then continuous 24 hours are taken as a slice, the entire first time series is divided into a plurality of day segments, if the predetermined unit time zone is a week, then continuous 7 days are taken as a slice, and for a month of first time series data are taken as a slice, a set of time series slices, namely, a first time series slice set, each representing the time series data in a specific unit time zone, is obtained.
In step S323, a first slice in the first time sequence slice set is extracted, where the first slice refers to a first concentration time sequence slice. Specifically, one slice is extracted from the first time sequence slice set as an example for further analysis and processing, the slice is called a first slice, the first slice is a part of a first intensive time sequence slice, the first slice is selected for subsequent model construction and data analysis demonstration, a processing method of the first slice can be generalized to the whole first time sequence slice set to realize unified processing and analysis of data of all time periods, the first slice is extracted from the first time sequence slice set in an index or random selection mode, if the first slice is processed in sequence, the first slice can be used as the first slice, and the first slice can be extracted randomly for analysis to verify the universality and the stability of the processing method.
Step S324, a first data set is constructed based on the first segment multi-domain feature information of the first intensity timing segment and the first intensity mode of the first intensity timing segment. Specifically, the first segment multi-domain feature information is a feature description of a first concentration time sequence segment, and comprises information of multiple aspects, namely, a time feature aspect, including a specific time range to which the segment belongs and classification of the time segment, a space feature aspect, including geographic position features of the region, building distribution and functions in the region, a historical trend feature aspect, analyzing the concentration change trend of the segment in similar time segments in historical data, an external factor feature aspect, including weather conditions, and sorting and quantifying the multi-domain feature information to form a feature vector or data set for subsequent data construction, the method comprises the steps of respectively calculating the mode of personnel density and vehicle density data in a first density time sequence segment, wherein the mode is a data value with the largest occurrence number in a group of data, reflecting the most common density level in the time period, similarly calculating the vehicle density, combining first segment multi-domain characteristic information and the first density mode to form a first data group, taking the multi-domain characteristic information as an input variable and the density mode as an output variable, and constructing a data pair form similar to (x, y), wherein x is a multi-domain characteristic information vector, y is a corresponding density mode, for example, x comprises a time feature, a space feature and an external factor feature, and y is the personnel density mode or the vehicle density mode of the time period.
And step S325, performing supervised learning on the first data set to obtain the intelligent prediction model. Specifically, a proper supervised learning algorithm is selected according to the characteristics of data and the requirements of a prediction task, for the problem of intensity prediction, if the data has a linear relation and features are relatively simple, a linear regression algorithm can be selected, if the data has a complex nonlinear relation and a plurality of feature variables, a neural network is selected, before the supervised learning is carried out, some preprocessing operations need to be carried out on a first data set, for example, the feature data are standardized or normalized, the numerical ranges of different features are on similar scales, so that the training efficiency and accuracy of a model are improved, for the problem of classification, if the features are text or category data, the encoding process needs to be carried out, the feature data are converted into a numerical form, the selected supervised learning algorithm is used for training the first data set, the feature information (x) in the first data set is used as input, the intensity mode (y) is used as a target output, the relation among the models is learned, the structures and parameters of the neural network are initialized, the data in the first data set are input into the neural network according to a batch, the forward propagation calculation of a prediction value, the numerical range between the prediction value and the actual error value is carried out on the basis of the similar scale, the model is well, the training condition of the model is reached, the specific error is reached, the condition of the model is reached, the intelligent model is reached, the training condition is reached after the specific time is reached, the model is reached, the iteration condition is reached, or the intelligent model is reached, and the training condition is reached.
In one possible implementation, step S324 further includes step S3241 of extracting a first vehicle-concentration time-series segment of the first-concentration time-series segment based on the first-segment multi-domain feature information of the first-concentration time-series segment and the first-concentration mode of the first-concentration time-series segment. Specifically, the first density time sequence segment is a subset of data of a time period divided from the first density time sequence according to a predetermined unit time zone, the subset including the density information of people and vehicles in the time period, the segment is a basis for further analyzing the characteristics of the vehicles and the people, for example, if the predetermined unit time zone is a day, the first density time sequence segment is data of the people and the vehicles in the day changing with time, and a data part specially related to the density of the vehicles, namely, the first vehicle density time sequence segment is separated from the first density time sequence segment. Because the vehicle concentration has its unique characteristics and rules, which need to be analyzed separately, extracted by identification or indexing of the data, assuming that the first time series segment of the concentration is stored in the form of a two-dimensional array or data table, one of which represents the person concentration and the other of which represents the vehicle concentration, the first time series segment of the vehicle concentration can be extracted by selecting the data corresponding to the vehicle concentration column.
And step S3242, performing time domain feature collection on the first vehicle concentration time sequence segment to obtain first vehicle segment time domain feature information. Specifically, the time domain features directly reflect the change condition of the vehicle concentration in the time dimension, the features can help us to know the aggregation and dispersion modes of the vehicles in different time periods, calculate the average value of the vehicle concentration data in the time sequence section of the first vehicle concentration, the average value reflects the overall level of the vehicle concentration in the time period, for example, if the average value of the vehicle concentration data in one day is higher, the vehicles are denser on the whole day, find out the maximum value in the time sequence section of the first vehicle concentration, namely the peak value of the vehicle concentration, the time point when the peak value appears may be the same as the traffic peak period, calculate the standard deviation of the vehicle concentration data, the standard deviation can measure the dispersion degree of the data, the standard deviation is larger, the change of the vehicle concentration in the time period is larger, the condition that the traffic flow fluctuation is larger is possible, observe the trend of the vehicle concentration changing along with time, the trend is gradually increased, gradually decreased or kept relatively stable, the trend is approximately described by fitting a trend line, for example, the linear regression fit time domain is used to represent the maximum value of the vehicle concentration, the time domain is used to represent the peak value, the time point when the vehicle concentration is likely to be the traffic trend, the feature information is integrated, and the feature information is formed.
Step S3243, frequency domain feature collection is performed on the first vehicle intensity spectrum obtained by the fast Fourier transform of the first vehicle intensity time sequence segment to obtain first vehicle segment frequency domain feature information. Specifically, a Fast Fourier Transform (FFT) converts a first vehicle concentration time-series segment of a time domain into a frequency domain, to obtain a first vehicle concentration frequency spectrum, in which frequency components of vehicle concentration data may be analyzed, so as to find some rules and features that are not obvious in the time domain, for example, certain periodic traffic patterns may be represented as peaks of specific frequencies in the frequency domain, the periodicity being related to daily commute rules, weekdays and weekend cycles, etc., and in the first vehicle concentration frequency spectrum, the frequency component with the highest energy, i.e., the dominant frequency, is found, and the dominant frequency reflects the most dominant periodic variation in the vehicle concentration data.
Step S3244, the first vehicle segment time domain feature information and the first vehicle segment frequency domain feature information form first vehicle segment multi-domain feature information. Specifically, the time domain feature information and the frequency domain feature information are combined to more fully describe the features of the time sequence segment of the first vehicle concentration, the time domain feature reflects the direct change of the vehicle concentration along with time, the frequency domain feature reveals the periodicity and the frequency component, the periodicity and the frequency component complement each other, richer information can be provided for subsequent analysis and model construction, the time domain feature information of the first vehicle segment and the frequency domain feature information of the first vehicle segment obtained before are combined to form a comprehensive feature set, namely the multi-domain feature information of the first vehicle segment, and the multi-domain feature information can be represented by a data structure, such as a vector or a structure body containing the time domain feature value and the frequency domain feature value.
Step S3245, obtaining the multi-domain characteristic information of the first person segment of the first person intensity time sequence segment. Specifically, for the personnel concentration part in the first concentration time sequence segment, an analysis method similar to the vehicle concentration is adopted, first the first personnel concentration time sequence segment is extracted, then time domain feature collection and frequency domain feature collection are respectively carried out, the time domain features comprise the mean value, the peak value, the valley value, the standard deviation, the trend and the like of the personnel concentration, the frequency domain features are obtained through fast Fourier transform and analysis of the personnel concentration time sequence segment, such as main frequency, frequency bandwidth and harmonic components and the like, the time domain features and the frequency domain features are integrated, and first personnel segment multi-domain feature information is obtained, wherein the representation mode of the first personnel segment multi-domain feature information is similar to that of the first vehicle segment multi-domain feature information, and the time domain feature can be a vector or a structure body containing various feature values.
Step S3246, the first vehicle segment multi-domain feature information and the first person segment multi-domain feature information form the first segment multi-domain feature information. Specifically, the first segment multi-domain feature information aims to comprehensively describe the overall situation of the first concentration time sequence segment, including the concentration features of two aspects of vehicles and personnel, the flows of the vehicles and the personnel are related to each other, the concentration situations of areas need to be analyzed and predicted accurately by considering at the same time, and the first vehicle segment multi-domain feature information and the first personnel segment multi-domain feature information are combined together to form a comprehensive feature set, namely the first segment multi-domain feature information.
And step S400, performing calibration analysis on the first prediction density according to the first dynamic characteristic information of the first space region to obtain a first actual density. Specifically, first dynamic characteristic information of a first space region is acquired through various channels, including sensor data, mobile device data, social media data and real-time event data, and data from different sources are integrated to form a data set. Then, the influence of the dynamic characteristic information on the first prediction concentration degree, such as the influence of traffic flow change and personnel flow change, is analyzed. Next, a calibration model or method is built. And inputting the first dynamic characteristic information into a calibration model or processing according to rules, adjusting the first prediction density to obtain the calibrated vehicle density and personnel density, and integrating the calibrated vehicle density and personnel density into a first actual density according to actual demands. And finally, verifying and evaluating the accuracy and rationality of the first actual concentration degree by comparing with the actual observation data and analyzing the time sequence, and if the error is large or does not accord with the change trend, adjusting and optimizing the whole process.
In one possible implementation manner, the calibration analysis is performed on the first predicted concentration according to the first dynamic feature information of the first spatial region to obtain a first actual concentration, and step S400 further includes step S410, where the first dynamic feature information is analyzed to obtain a first dynamic concentration time sequence. The first dynamic characteristic information comprises various data about real-time change of a first space area, and covers various aspects of information, such as real-time flowing conditions of people and vehicles, occurrence of real-time change of traffic conditions and the like, the current dynamic state of the first space area can be reflected continuously and updated along with time, the real-time flowing conditions of the vehicles can be obtained through traffic monitoring equipment, such as the number of vehicles, the speed, the road occupancy and the like, the change of the traffic conditions comprises whether the road is congested, the increase and decrease of traffic flow and the like, information related to the people and the vehicle concentration is extracted from the first dynamic characteristic information, the people concentration in different time points and different areas is calculated according to the data of a people flow sensor, the vehicle concentration in different road sections and time periods is calculated according to the data of the traffic monitoring equipment, the people and the vehicle concentration information which change along with time are integrated together to form a first dynamic concentration time sequence, and the time sequence records the dynamic change conditions of the people and the vehicle concentration along with time and is a time sequence of data sequence.
Step S420, performing trend line analysis on the first dynamic concentration time sequence based on the spectrum density function principle to obtain a first future concentration at a first future time. In particular, the spectral density function is used to analyze the frequency characteristics of the time series data, identifying periodic and trending components in the data. The periodic law can reflect changes such as peak time for the personnel concentration time sequence, and the like for the vehicle concentration time sequence has similar peak law. And (3) selecting a linear regression model as a trend line analysis method through fitting the spectral density function result. The linear regression obtains a best fit line by fitting the historical intensity data, and predicts using the line to obtain a first future intensity. The specific prediction time point is determined according to actual requirements and data resolution, and the future time point is substituted into a fitted linear regression equation to calculate corresponding personnel and vehicle concentration values, so that the first future concentration is obtained.
And step S430, adjusting the first prediction density according to a first corresponding relation between the first future time and the first future density to obtain the first actual density. Specifically, since the first prediction concentration is obtained based on historical data and model prediction, the first future concentration obtained by trend line analysis is a future value based on the current dynamic change trend prediction of the first spatial region, and in order to obtain a more accurate first actual concentration, a weighted average method is needed to determine a weight coefficient(1) For balancing contributions of the first predicted concentration and the first future concentration, e.g. if=0.7, Then indicates a greater propensity to believe the first prediction concentration, an=0.3, Which represents a more important first future concentration based on trend line analysis, assuming that the first predicted concentration is(Including predicted personnel and vehicle concentration), the first future concentration is(Also including future person and vehicle concentration), for the adjustment of the actual person concentration, the calculation is that the actual person concentration =. Wherein the method comprises the steps ofIs the first predicted concentration of people,Is the first future personnel concentration, and for the adjustment of the actual vehicle concentration, the calculation formula is that the actual vehicle concentration =. Wherein the method comprises the steps ofIs the first predicted concentration of people,The method is characterized in that the first future personnel concentration is obtained by fusing the first prediction concentration with the first future concentration obtained based on trend line analysis through the calculation, so that the first actual concentration which is more in line with the actual condition of the first space region is obtained, the first actual concentration not only considers the prediction results of historical data and models, but also combines the current dynamic change trend to predict the future.
And S500, when the first actual concentration is not in a preset concentration threshold, performing treatment scheduling adjustment on the first space region. Specifically, first, a predetermined concentration threshold is determined, which is set by analyzing historical data, expert evaluation and city development planning by comprehensively considering the functional characteristics, resource bearing capacity, city planning targets and other factors of the first space region, and different functional regions have different standards. And then evaluating the first actual concentration, analyzing the dynamic characteristic information of the area to reflect the actual congestion and flow conditions, and measuring the actual concentration of personnel and vehicles in various modes and comparing the actual concentration with a preset threshold. And when the first actual concentration is not in the threshold, performing scheduling adjustment. In traffic management, traffic flow control is included, such as public transportation optimization by dynamically adjusting signal lamp time. In the aspect of public facility configuration, a parking space is planned, and public service facilities are adjusted. In the aspect of personnel evacuation guiding, guiding identification and information release are set, and emergency drilling and training are also carried out so as to enhance personnel coping capacity and safety awareness.
Hereinabove, a city governance scheduling method based on dynamic evaluation according to an embodiment of the present invention is described in detail with reference to fig. 1. Next, a dynamic evaluation-based urban governance scheduling system according to an embodiment of the present invention will be described with reference to fig. 2.
The urban governance scheduling system based on dynamic evaluation is used for solving the technical problem that the prior art cannot respond to the dynamic changes of the environment and the personnel flow in time, and achieves the technical effect of analyzing the environment and the personnel flow dynamically in real time. A city governance scheduling system based on dynamic evaluation comprises a first space region acquisition module 10, a first governance fitness acquisition module 20, a first prediction concentration acquisition module 30, a first actual concentration acquisition module 40 and a governance scheduling adjustment module 50.
The first spatial region obtaining module 10 is configured to obtain a first spatial region, where the first spatial region refers to any region in a spatial region set obtained by dividing a target city, and the first spatial region has an identifier of a first functional characteristic.
The first governance fitness obtaining module 20 is configured to analyze the collected first region feature information of the first spatial region to obtain a first governance fitness of the first spatial region determined to be the first functional characteristic.
The first prediction density acquisition module 30 is configured to call a density database to perform a density prediction analysis on the first spatial region when the first treatment fitness meets a predetermined fitness threshold, so as to obtain a first prediction density.
The first actual concentration obtaining module 40 is configured to perform calibration analysis on the first predicted concentration according to the first dynamic feature information of the first spatial region, so as to obtain a first actual concentration.
The governance schedule adjustment module 50 is configured to perform governance schedule adjustment on the first spatial region when the first actual concentration is not within a predetermined concentration threshold.
Next, the specific configuration of the first governance fitness acquiring module 20 will be described in detail. As described above, the first region feature information of the collected first spatial region is analyzed to obtain a first governance fitness of the first spatial region determined to be the first functional characteristic, and the first governance fitness acquisition module 20 further comprises a predetermined topography and topography index reading unit for reading a predetermined topography and topography index, a first topography and topography feature information acquisition unit for performing multidimensional feature collection on the first spatial region based on the predetermined topography and topography index to obtain first topography and topography feature information, a predetermined flow index reading unit for reading a predetermined flow index, a multidimensional feature collection unit for performing multidimensional feature collection on the first spatial region based on the predetermined flow index to obtain first flow feature information, and a first governance fitness acquisition unit for analyzing the first topography and topography feature information and the first flow fitness feature information.
The first treatment fitness obtaining unit further comprises a city treatment database calling subunit, a first historical treatment record, a second historical treatment record and a second historical treatment record, wherein the city treatment database calling subunit is used for calling a city treatment database and extracting a first historical treatment record in the city treatment database, and the first historical treatment record refers to the city treatment record of the first functional characteristic of scheduling the first historical space region treatment; the system comprises a first historical topography feature information constructing subunit for constructing first historical topography feature information of the first historical space region based on the preset topography index, a first situation similarity acquiring subunit for carrying out similarity analysis on the first topography feature information and the first historical topography feature information to obtain first situation similarity, a first historical treatment difficulty extracting subunit for extracting first historical treatment difficulty in the first historical treatment record when the first situation similarity accords with a preset similarity threshold, a first treatment situation coupling degree acquiring subunit for taking the reciprocal of the first historical treatment difficulty as first treatment situation coupling degree, a variation weighting processing subunit for carrying out variation weighting processing on a plurality of flow feature parameters in the first flow feature information, the system comprises a first governance flow coupling degree acquisition subunit, a first governance fitness acquisition subunit, a second governance flow coupling degree generation subunit and a second governance fitness acquisition subunit, wherein the first governance fitness acquisition subunit is used for taking the average value of the first governance situation coupling degree and the first governance flow coupling degree as the first governance fitness.
Next, the specific configuration of the first prediction density acquisition module 30 will be described in detail. When the first governance fitness meets a predetermined fitness threshold, the first governance fitness is subjected to a density prediction analysis on the first space region to obtain a first prediction density, the first prediction density acquisition module 30 further comprises a first density time sequence extraction unit, the first density time sequence extraction unit is used for extracting a first density time sequence in the density database, the first density database comprises historical density record data of a plurality of historical space regions consistent with the first functional characteristic, the first density time sequence comprises a first vehicle density time sequence and a first person density time sequence, the intelligent prediction model activation unit is used for activating an intelligent prediction model, the intelligent prediction model comprises a vehicle prediction layer and a person prediction layer, the first prediction vehicle density acquisition unit is used for extracting first density time sequences in the density database, the first density time sequence extraction unit comprises a plurality of first person density prediction units, the first person density prediction units are used for analyzing the first person density information of the first vehicle density time sequence, the first person density information is formed by the first person density prediction units, and the first person density information is analyzed by the first person density prediction units, and the first person density information is formed by the first person density prediction units.
The intelligent prediction model activation unit further comprises a preset unit time zone reading subunit, a first time slice set acquisition subunit and a data supervision learning subunit, wherein the preset unit time zone reading subunit is used for reading a preset unit time zone, the first time slice set acquisition subunit is used for carrying out slicing processing on the first concentration time sequence by taking the preset unit time zone as constraint to obtain a first time slice set, the first slice extraction subunit is used for extracting a first slice in the first time slice set, the first slice refers to a first concentration time sequence segment, the first data set construction subunit is used for constructing a first data set based on first segment multi-domain characteristic information of the first concentration time sequence segment and a first concentration time sequence segment, and the data supervision learning subunit is used for carrying out supervision learning on the first data set to obtain the intelligent prediction model.
The first data set is composed based on first segment multi-domain feature information of the first concentration time sequence segment and a first concentration mode of the first concentration time sequence segment, and the first data set composing subunit further comprises a concentration time sequence segment extraction micro-unit used for extracting the first vehicle concentration time sequence segment in the first concentration time sequence segment, a time domain feature collection micro-unit used for collecting time domain features of the first vehicle concentration time sequence segment to obtain first vehicle segment time domain feature information, a frequency domain feature collection micro-unit used for collecting frequency domain features of the first vehicle segment frequency domain spectrum obtained by fast Fourier transformation of the first vehicle concentration time sequence segment to obtain first vehicle segment frequency domain feature information, a first vehicle segment multi-domain feature information composing micro-unit used for extracting the first vehicle concentration time sequence segment in the first concentration time sequence segment, a time domain feature collection micro-unit used for collecting time domain features of the first vehicle segment to obtain the first vehicle segment time domain feature information, and a frequency domain feature collection micro-unit used for fast Fourier transformation of the first vehicle concentration time sequence segment obtained by the first vehicle concentration time sequence segment, and a first person multi-domain feature information composing micro-unit used for obtaining the first person segment feature information.
Next, the specific configuration of the first actual density acquisition module 40 will be described in detail. As described above, the first prediction density is calibrated and analyzed according to the first dynamic characteristic information of the first spatial region to obtain a first actual density, and the first actual density acquisition module 40 further includes a first dynamic density time sequence analysis unit for analyzing the first dynamic characteristic information to obtain a first dynamic density time sequence, a first future density acquisition unit for performing trend line analysis on the first dynamic density time sequence based on a spectral density function principle to obtain a first future density at a first future time, and a first actual density acquisition unit for adjusting the first prediction density according to a first correspondence between the first future time and the first future density to obtain the first actual density.
The urban governance scheduling system based on dynamic evaluation provided by the embodiment of the invention can execute the urban governance scheduling method based on dynamic evaluation provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.
Although the present application makes various references to certain modules in a system according to an embodiment of the present application, any number of different modules may be used and run on a user terminal and/or a server, and each unit and module included are merely divided according to functional logic, but are not limited to the above-described division, so long as the corresponding functions can be implemented, and in addition, specific names of each functional unit are only for convenience of distinguishing from each other, and are not intended to limit the scope of protection of the present application.
The above embodiments do not limit the scope of the present application. It will be apparent to those skilled in the art that various modifications, combinations, and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of the present application. In some cases, the acts or steps recited in the present application may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.