WO2022267870A1 - Index optimization method, server, and computer-readable storage medium - Google Patents

Abstract

An index optimization method, a server, and a computer-readable storage medium. A method used for a first server comprises: obtaining a first optimization result and a service feature that are encrypted and sent respectively by each second server, wherein the first optimization result is determined by a first log information, the first log information carries an operation feature in a process in which a first performance index is restored from an abnormal state to a normal state, the service feature is used for characterizing environment information that executes the operation feature, and the first optimization result comprises a first optimization model for the first performance index; constructing a federated learning model according to all first optimization results and all service features, and determining a second optimization model according to the federated learning model; and respectively encrypting and sending a second optimization model to each second server.

Description

Translated fromChinese

指标优化方法及服务器、计算机可读存储介质Index optimization method, server, and computer-readable storage medium

相关申请的交叉引用Cross References to Related Applications

本申请基于申请号为202110707644.4、申请日为2021年6月24日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。This application is based on a Chinese patent application with application number 202110707644.4 and a filing date of June 24, 2021, and claims the priority of this Chinese patent application. The entire content of this Chinese patent application is hereby incorporated by reference into this application.

技术领域technical field

本申请实施例涉及但不限于人工智能技术领域，尤其涉及一种指标优化方法及服务器、计算机可读存储介质。The embodiments of the present application relate to but are not limited to the technical field of artificial intelligence, and in particular relate to an index optimization method, a server, and a computer-readable storage medium.

背景技术Background technique

随着社会不断发展，目前各国陆续出台了一系列数据保护法律法规，旨在保护公民与企业数据隐私，可见对用户数据的保护以及数据的安全性需求将成为未来的不可逆趋势。由于不同地区的数据不能随意转移，因此这些法律法规的建立在不同程度上对人工智能传统的数据处理模式提出了新的挑战。目前，电信运营商的运维人员对网管系统需要持续的观察关键的性能指标，并进行相关性能指标的优化，通常而言，影响性能指标的操作有很多，例如，规避内部或外部干扰、修改邻区配置以及修改工程参数配置等，这些操作都会记录在日志信息中，相关技术中的性能指标优化，即根据这些操作来构造一套算法模型以进行指标优化，但是实际优化效果很一般，性能指标优化成功率相对较低。With the continuous development of society, countries have successively issued a series of data protection laws and regulations to protect the data privacy of citizens and enterprises. It can be seen that the protection of user data and the demand for data security will become an irreversible trend in the future. Since data in different regions cannot be transferred at will, the establishment of these laws and regulations pose new challenges to the traditional data processing mode of artificial intelligence to varying degrees. At present, the operation and maintenance personnel of telecom operators need to continuously observe key performance indicators of the network management system and optimize related performance indicators. Generally speaking, there are many operations that affect performance indicators, such as avoiding internal or external interference, modifying Neighboring cell configuration and modification of project parameter configuration, etc., these operations will be recorded in the log information, performance index optimization in related technologies, that is, to construct a set of algorithm models based on these operations for index optimization, but the actual optimization effect is very general, performance The index optimization success rate is relatively low.

发明内容Contents of the invention

以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。The following is an overview of the topics described in detail in this article. This summary is not intended to limit the scope of the claims.

本申请实施例提供了一种指标优化方法及服务器、计算机可读存储介质，能够提高性能指标优化成功率。Embodiments of the present application provide an index optimization method, a server, and a computer-readable storage medium, which can improve the success rate of performance index optimization.

第一方面，本申请实施例提供了一种指标优化方法，应用于第一服务器，所述方法包括：获取由各个第二服务器分别加密发送的与第一性能指标对应的第一优化结果和业务特征，其中，所述第一优化结果为由所述第二服务器的第一日志信息确定，所述第一日志信息携带有所述第一性能指标从异常状态恢复到正常状态的过程中的操作特征，所述业务特征用于表征执行所述操作特征的环境信息，所述第一优化结果包括针对所述第一性能指标的第一优化模型；以所有所述第一优化结果和所有所述业务特征作为输入数据而构建联邦学习模型，并根据所述联邦学习模型确定第二优化模型；向各个所述第二服务器分别加密发送所述第二优化模型，以使各个所述第二服务器根据所述第二优化模型更新所述第一优化模型。In the first aspect, the embodiment of the present application provides an index optimization method, which is applied to the first server, and the method includes: obtaining the first optimization results corresponding to the first performance index and the business feature, wherein, the first optimization result is determined by the first log information of the second server, and the first log information carries the operation in the process of restoring the first performance indicator from an abnormal state to a normal state feature, the business feature is used to characterize the environment information for performing the operation feature, the first optimization result includes a first optimization model for the first performance index; with all the first optimization results and all the Constructing a federated learning model based on business characteristics as input data, and determining a second optimization model according to the federated learning model; sending the second optimized model to each of the second servers in encrypted form, so that each of the second servers according to The second optimization model updates the first optimization model.

第二方面，本申请实施例还提供了一种指标优化方法，应用于第二服务器，所述第二服务器设置有多个，所述方法包括：获取第一日志信息和业务特征，所述第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，所述业务特征用于表征执行所述操作特征的环境信息；根据所述操作特征确定与所述第一性能指标对应的第一优化结果，所述第一优化结果包括针对所述第一性能指标的第一优化模型；向第一服务器加密传输所述 第一优化结果和所述业务特征，以使所述第一服务器根据联邦学习模型确定第二优化模型，所述联邦学习模型为由所述第一服务器以多个所述第二服务器分别发送的所述第一优化结果和所述业务特征作为输入数据而构建；接收由所述第一服务器加密发送的所述第二优化模型，并根据所述第二优化模型更新所述第一优化模型。In the second aspect, the embodiment of the present application also provides an index optimization method, which is applied to the second server, and there are multiple second servers. The method includes: obtaining the first log information and business characteristics, and the second A log information carries operating characteristics of the first performance index in the process of returning from an abnormal state to a normal state, and the business characteristics are used to characterize the environment information for executing the operating characteristics; A first optimization result corresponding to the performance index, the first optimization result including a first optimization model for the first performance index; encrypting and transmitting the first optimization result and the service characteristics to the first server, so that the The first server determines a second optimization model according to a federated learning model, and the federated learning model uses the first optimization result and the business characteristics respectively sent by a plurality of second servers as input by the first server data; receiving the second optimization model encrypted and sent by the first server, and updating the first optimization model according to the second optimization model.

第三方面，本申请实施例还提供了一种服务器，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如上所述第一方面的指标优化方法。In the third aspect, the embodiment of the present application also provides a server, including: a memory, a processor, and a computer program stored on the memory and operable on the processor, and the processor implements the above-mentioned server when executing the computer program. Describe the index optimization method in the first aspect.

第四方面，本申请实施例还提供了一种服务器，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如上所述第二方面的指标优化方法。In the fourth aspect, the embodiment of the present application also provides a server, including: a memory, a processor, and a computer program stored on the memory and operable on the processor, and the processor implements the above-mentioned computer program when executing the computer program. Describe the index optimization method in the second aspect.

第五方面，本申请实施例还提供了一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行如上所述第一方面的指标优化方法，或者，执行如上所述第二方面的指标优化方法。In the fifth aspect, the embodiment of the present application also provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to execute the index optimization method in the first aspect as described above, or to execute the above-mentioned The index optimization method of the second aspect.

本申请的其它特征和优点将在随后的说明书中阐述，并且，部分地从说明书中变得显而易见，或者通过实施本申请而了解。本申请的目的和其他优点可通过在说明书、权利要求书以及附图中所特别指出的结构来实现和获得。Additional features and advantages of the application will be set forth in the description which follows, and, in part, will be obvious from the description, or may be learned by practice of the application. The objectives and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

附图说明Description of drawings

附图用来提供对本申请技术方案的进一步理解，并且构成说明书的一部分，与本申请的实施例一起用于解释本申请的技术方案，并不构成对本申请技术方案的限制。The accompanying drawings are used to provide a further understanding of the technical solution of the present application, and constitute a part of the specification, and are used together with the embodiments of the present application to explain the technical solution of the present application, and do not constitute a limitation to the technical solution of the present application.

图1是本申请一个实施例提供的用于执行指标优化方法的网络拓扑的示意图；FIG. 1 is a schematic diagram of a network topology for performing an index optimization method provided by an embodiment of the present application;

图2是本申请一个实施例提供的指标优化方法的流程图；Fig. 2 is a flowchart of an index optimization method provided by an embodiment of the present application;

图3是本申请一个实施例提供的操作特征的示意图；Fig. 3 is a schematic diagram of the operating features provided by an embodiment of the present application;

图4是本申请一个实施例提供的业务特征的示意图；Fig. 4 is a schematic diagram of service features provided by an embodiment of the present application;

图5是本申请另一个实施例提供的指标优化方法的流程图；Fig. 5 is a flowchart of an index optimization method provided by another embodiment of the present application;

图6是本申请一个实施例提供的指标优化方法中确定第一优化结果的流程图；Fig. 6 is a flow chart of determining the first optimization result in the index optimization method provided by an embodiment of the present application;

图7是本申请一个实施例提供的指标优化方法中传输第一优化结果和业务特征的流程图；Fig. 7 is a flow chart of transmitting the first optimization result and service characteristics in the indicator optimization method provided by an embodiment of the present application;

图8是本申请一个实施例提供的服务器的示意图；Fig. 8 is a schematic diagram of a server provided by an embodiment of the present application;

图9是本申请另一个实施例提供的服务器的示意图。Fig. 9 is a schematic diagram of a server provided by another embodiment of the present application.

具体实施方式detailed description

为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请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申请，并不用于限定本申请。In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, not to limit the present application.

需要注意的是，虽然在装置示意图中进行了功能模块划分，在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于装置中的模块划分，或流程图中的顺序执行所示出或描述的步骤。说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。It should be noted that although the functional modules are divided in the schematic diagram of the device, and the logical sequence is shown in the flowchart, in some cases, it can be executed in a different order than the module division in the device or the flowchart in the flowchart. steps shown or described. The terms "first", "second" and the like in the specification and claims and the above drawings are used to distinguish similar objects, and not necessarily used to describe a specific sequence or sequence.

本申请提供了一种指标优化方法及服务器、计算机可读存储介质，通过获取由各个第二 服务器发送的第一优化结果和业务特征，使得在确定的第一优化结果的基础上结合业务特征而构建联邦学习模型，从而基于联邦学习模型可以得到相比第一优化模型具有更高精确度的第二优化模型，相比于相关技术，能够在不泄露或转移数据的条件下进行性能指标优化，安全性更高，并通过向各个第二服务器分别发送第二优化模型，使得各个第二服务器根据第二优化模型更新第一优化模型，从而能够提高第二服务器的性能指标优化成功率。The present application provides an index optimization method, a server, and a computer-readable storage medium. By obtaining the first optimization results and business characteristics sent by each second server, the business characteristics can be determined on the basis of the determined first optimization results. Construct a federated learning model, so that based on the federated learning model, a second optimization model with higher accuracy than the first optimization model can be obtained. Compared with related technologies, performance index optimization can be performed without leaking or transferring data. The security is higher, and by sending the second optimization model to each second server respectively, each second server updates the first optimization model according to the second optimization model, so that the success rate of performance index optimization of the second server can be improved.

下面结合附图，对本申请实施例作进一步阐述。The embodiments of the present application will be further described below in conjunction with the accompanying drawings.

如图1所示，图1是本申请一个实施例提供的用于执行指标优化方法的网络拓扑的示意图。As shown in FIG. 1 , FIG. 1 is a schematic diagram of a network topology for performing an index optimization method provided by an embodiment of the present application.

在图1的示例中，该网络拓扑包括但不限于：第一服务器100和第二服务器200，其中，第一服务器100可以设置为一个，第二服务器200可以设置为多个，每个第二服务器200可以设置在不同的国家、地区以及区域等，且每个第二服务器200与第一服务器100之间可以通过内网交换机300进行连接，该网络可以是无线网络，例如，在基站条件下的无线通信网络，也可以是有线网络等，这并未限制。In the example of FIG. 1, the network topology includes but is not limited to: afirst server 100 and asecond server 200, wherein thefirst server 100 can be set to one, and thesecond server 200 can be set to multiple, each second Theserver 200 can be set in different countries, regions and regions, etc., and eachsecond server 200 can be connected to thefirst server 100 through anintranet switch 300. The network can be a wireless network, for example, under the condition of a base station The wireless communication network can also be a wired network, etc., which is not limited.

在一实施例中，第二服务器200可以但不限于用于监控用户数据以及对用户数据中的相关指标进行性能优化，其中，用户数据的来源并不限定，理论上第二服务器200只要能够获取到用户数据即可，例如，参照图1，每个第二服务器200均通过相应的交换机与地区网管系统400连接，即，每个第二服务器200能够从相应的地区网管系统400获取第一日志信息，从而根据第一日志信息得到相应的地区网管系统400的用户数据，因此，通过第二服务器200可以实现网管性能指标优化，或者，不局限于地区网管系统400，若第二服务器200所在区域的用户数据存在安全、隐私或异常问题，则第二服务器200也可以对该用户数据进行性能优化，换言之，第二服务器200所获取的用户数据的来源可以不限制。In an embodiment, thesecond server 200 can be used, but not limited to, to monitor user data and perform performance optimization on related indicators in the user data. The source of the user data is not limited. In theory, as long as thesecond server 200 can obtain For example, with reference to FIG. 1, eachsecond server 200 is connected to the regionalnetwork management system 400 through a corresponding switch, that is, eachsecond server 200 can obtain the first log from the corresponding regionalnetwork management system 400 Information, so as to obtain the user data of the corresponding regionalnetwork management system 400 according to the first log information, therefore, the network management performance index optimization can be realized through thesecond server 200, or, not limited to the regionalnetwork management system 400, if thesecond server 200 is located in the area If the user data has security, privacy or abnormal issues, thesecond server 200 may also perform performance optimization on the user data. In other words, the source of the user data acquired by thesecond server 200 may not be limited.

在一实施例中，第一服务器100作为联邦学习的核心，可以通过加密机制下的参数交换方式，即，在不违反数据隐私法规的情况下，基于从各个第二服务器200获取的数据信息建立一个虚拟的共有模型，相应地，联邦学习的置信度模型则部署在各个第二服务器200上作为工作节点，各个第二服务器200根据联邦学习的置信度模型进行性能指标优化，在这个过程中，不会泄露任何私有数据，具有较高的安全性，其中，联邦学习(Federated Learn ing)是一个机器学习框架，能有效帮助多个机构在满足用户隐私保护、数据安全和政府法规的要求下，进行数据使用和机器学习建模,以一个基本例子简要说明其功能原理：In an embodiment, thefirst server 100, as the core of federated learning, can exchange parameters under the encryption mechanism, that is, without violating data privacy regulations, establish A virtual common model. Correspondingly, the confidence model of federated learning is deployed on eachsecond server 200 as a working node, and eachsecond server 200 optimizes the performance index according to the confidence model of federated learning. In this process, It will not leak any private data and has high security. Among them, Federated Learning is a machine learning framework that can effectively help multiple organizations meet the requirements of user privacy protection, data security and government regulations. Carry out data usage and machine learning modeling, and briefly illustrate its functional principle with a basic example:

假设有两个不同且不相关的企业1和企业2,它们拥有不同数据，比如,企业1配置有用户特征数据，企业2配置有产品特征数据和标注数据，根据相关数据法规，这两个企业各自的数据不能简单地直接加以合并。考虑解决上述问题，一种手段为双方各自建立一个任务模型,每个任务模型可以是分类模型或预测模型,而这些任务模型可以在获得数据时受到各自企业的用户的认可,但出现了新的问题，即，如何在企业1和企业2各自一侧分别建立高质量的模型。采用联邦学习系统即可以解决这个问题，在各个企业的自有数据不出本地的情况下,联邦学习系统可以通过加密机制下的参数交换方式,即在不违反数据隐私法规情况下,建立一个虚拟的共有模型，该虚拟模型相当于将企业1和企业2各自的数据聚合在一起，但是在建立虚拟模型的时候,企业1和企业2各自的数据本身不移动,因此并不会泄露隐私和影响数据合规,在这样的联邦学习机制下，建好的模型在企业1和企业2各自的区域仅为本地的目标服务，从而确保每个企业的数据在使用时不会受到影响。Assume that there are two different and irrelevant enterprises 1 and 2, which have different data. For example, enterprise 1 is configured with user characteristic data, and enterprise 2 is configured with product characteristic data and annotation data. According to relevant data regulations, the two enterprises The respective data cannot simply be merged directly. Considering to solve the above problems, one method is to establish a task model for both parties. Each task model can be a classification model or a prediction model, and these task models can be recognized by users of their respective enterprises when data is obtained, but new The problem is, how to build high-quality models on each side of enterprise 1 and enterprise 2 respectively. This problem can be solved by adopting the federated learning system. In the case that each enterprise’s own data does not go out of the local area, the federated learning system can establish a virtual This virtual model is equivalent to aggregating the respective data of enterprise 1 and enterprise 2, but when establishing the virtual model, the respective data of enterprise 1 and enterprise 2 do not move, so the privacy and influence will not be leaked Data compliance, under such a federated learning mechanism, the built models only serve local targets in the respective regions of enterprise 1 and enterprise 2, so as to ensure that the data of each enterprise will not be affected when used.

在一实施例中，用户数据即终端数据，其中，终端可以设置为多个，各个终端均可以称为接入终端、用户设备(User Equipment，UE)、用户单元、用户站、移动站、移动台、远方站、远程终端、移动设备、用户终端、无线通信设备、用户代理或用户装置。例如，各个终端均可以是蜂窝电话、无绳电话、会话启动协议(Session Initiation Protocol，SIP)电话、无线本地环路(Wireless Local Loop，WLL)站、个人数字处理(Personal Digital Assistant，PDA)、具有无线通信功能的手持设备、计算设备或连接到无线调制解调器的其它处理设备、车载设备、可穿戴设备、5G网络或者未来5G以上网络中的终端设备等，本实施例对此并不作具体限定。In one embodiment, the user data is terminal data, where multiple terminals can be set, and each terminal can be called an access terminal, user equipment (User Equipment, UE), user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, wireless communication device, user agent or user device. For example, each terminal may be a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA), a Handheld devices with wireless communication functions, computing devices or other processing devices connected to wireless modems, vehicle-mounted devices, wearable devices, 5G networks or terminal devices in future 5G or higher networks, etc., are not specifically limited in this embodiment.

第一服务器100和第二服务器200均可以分别包括有存储器和处理器，其中，存储器和处理器可以通过总线或者其他方式连接。Both thefirst server 100 and thesecond server 200 may respectively include a memory and a processor, where the memory and the processor may be connected through a bus or in other ways.

存储器作为一种非暂态计算机可读存储介质，可用于存储非暂态软件程序以及非暂态性计算机可执行程序。此外，存储器可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施方式中，存储器可选包括相对于处理器远程设置的存储器，这些远程存储器可以通过网络连接至该处理器。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。As a non-transitory computer-readable storage medium, memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes memory located remotely from the processor, and these remote memories may be connected to the processor via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

本申请实施例描述的网络拓扑以及应用场景是为了更加清楚的说明本申请实施例的技术方案，并不构成对于本申请实施例提供的技术方案的限定，本领域技术人员可知，随着网络拓扑的演变和新应用场景的出现，本申请实施例提供的技术方案对于类似的技术问题，同样适用。The network topology and application scenarios described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute limitations on the technical solutions provided by the embodiments of the present application. Those skilled in the art know that with the network topology The evolution of the technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

本领域技术人员可以理解的是，图1中示出的网络拓扑并不构成对本申请实施例的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。Those skilled in the art can understand that the network topology shown in Figure 1 does not constitute a limitation to the embodiment of the present application, and may include more or less components than shown in the figure, or combine some components, or different components layout.

在图1所示的网络拓扑中，第一服务器100和第二服务器200可以分别调用其储存的性能指标优化程序，以执行指标优化方法。In the network topology shown in FIG. 1 , thefirst server 100 and thesecond server 200 can respectively call their stored performance index optimization programs to execute the index optimization method.

基于上述网络拓扑的结构，提出本申请的指标优化方法的各个实施例。Based on the structure of the above network topology, various embodiments of the index optimization method of the present application are proposed.

如图2所示，图2是本申请一个实施例提供的指标优化方法的流程图，可以但不限于应用于如图1实施例所示的网络拓扑中的第一服务器，该指标优化方法包括但不限于步骤S100至S300。As shown in Figure 2, Figure 2 is a flow chart of an index optimization method provided by an embodiment of the present application, which can be but not limited to be applied to the first server in the network topology shown in the embodiment of Figure 1, the index optimization method includes But not limited to steps S100 to S300.

步骤S100:获取由各个第二服务器分别加密发送的与第一性能指标对应的第一优化结果和业务特征，其中，第一优化结果为由第二服务器的第一日志信息确定，第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，业务特征用于表征执行操作特征的环境信息，第一优化结果包括针对第一性能指标的第一优化模型。Step S100: Obtain the first optimization results and business characteristics corresponding to the first performance indicators encrypted and sent by each second server, wherein the first optimization results are determined by the first log information of the second server, and the first log information Carrying the operating characteristics of the first performance index in the process of returning from the abnormal state to the normal state, the business characteristics are used to characterize the environment information for executing the operating characteristics, and the first optimization result includes the first optimization model for the first performance index.

在一实施例中，通过获取由各个第二服务器发送的第一优化结果和业务特征，以之作为构建联邦学习模型的基础数据，以便于在后续的步骤中实现构建联邦学习模型。In an embodiment, the first optimization results and service characteristics sent by each second server are obtained as basic data for constructing a federated learning model, so as to realize building a federated learning model in subsequent steps.

需要说明的是，第一性能指标不是唯一确定的，可以根据实际场景进行选择，例如，在网络参数配置场景下，想要了解网络配置情况，因此相对较为关注网络邻区情况，在这种情景下，则确定网络邻区指标为第一性能指标，进而对网络邻区指标进行进一步地分析处理；第一日志信息可以由第二服务器实时获取，也可以由第二服务器预先获取并进行存储，根据前述关于网络拓扑的实施例可知，第一日志信息的载体可以为多种，例如，可以是网管系统、业务系统等，这并未限制。It should be noted that the first performance index is not uniquely determined, and can be selected according to the actual scenario. For example, in the scenario of network parameter configuration, one wants to know the configuration of the network, so relatively more attention is paid to the condition of the network neighbors. In this scenario Next, the network neighbor index is determined as the first performance index, and then the network neighbor index is further analyzed and processed; the first log information can be obtained by the second server in real time, or can be obtained and stored in advance by the second server, According to the above-mentioned embodiments about the network topology, it can be known that the carrier of the first log information may be various, for example, it may be a network management system, a service system, etc., which is not limited.

在一实施例中，第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，由于该操作特征伴随着第一性能指标的状态变更过程，因此，可以认为该操作特征对于第一性能指标的优化有影响，因此，可以基于该操作特征来进一步地确认与第一性能指标对应的第一优化结果。In one embodiment, the first log information carries the operating characteristics of the first performance indicator in the process of recovering from the abnormal state to the normal state. Since the operating characteristics are accompanied by the state change process of the first performance indicator, it can be considered that the The operating characteristic has an influence on the optimization of the first performance index, therefore, the first optimization result corresponding to the first performance index can be further confirmed based on the operating characteristic.

可以理解地是，操作特征可以对应于第一性能指标从异常状态恢复到正常状态的过程中的全部操作特征，也可以对应于第一性能指标从异常状态恢复到正常状态的过程中的部分操作特征，这并未限制。It can be understood that the operating characteristics may correspond to all operating characteristics in the process of restoring the first performance index from an abnormal state to a normal state, or may correspond to some operations in the process of restoring the first performance index from an abnormal state to a normal state. features, this is not limited.

在一实施例中，如图3所示，第一日志信息以存储列表形式保存有操作特征的记录，操作特征包括如下类型中的至少一种：与第一性能指标对应的网元，记为Neid；与第一性能指标对应的对象，记为Object；针对第一性能指标执行操作的名称，记为Operation Name；第一性能指标从异常状态恢复到正常状态前的数值,记为Before Value；第一性能指标从异常状态恢复到正常状态后的数值,记为Modify Value；针对第一性能指标执行操作的结果,记为Result。In one embodiment, as shown in FIG. 3 , the first log information stores records of operation characteristics in the form of a storage list, and the operation characteristics include at least one of the following types: the network element corresponding to the first performance index, denoted as Neid; the object corresponding to the first performance index is recorded as Object; the name of the operation performed on the first performance index is recorded as Operation Name; the value before the first performance index returns from abnormal state to normal state is recorded as Before Value; The value of the first performance index after returning from the abnormal state to the normal state is recorded as Modify Value; the result of performing operations on the first performance index is recorded as Result.

需要说明的是，在上述操作特征的各种类型中，网元和对象体现了第一性能指标的作用目标，针对第一性能指标执行操作的名称和结果体现了操作特征的具体含义，而第一性能指标从异常状态恢复到正常状态之前和之后的数值可以体现操作特征的作用程度，因此，通过操作特征可以表征对于第一性能指标造成的操作影响，因此，第一优化结果可以基于操作特征来进一步地确定，显然地，在不同的应用场景下，针对不同的性能指标，所对应的操作特征也可能是不同的，本实施例并未对其限制。It should be noted that among the various types of the above operating characteristics, network elements and objects embody the target of the first performance index, and the names and results of operations performed on the first performance index embody the specific meaning of the operating characteristics, while the first performance index The value of a performance index before and after returning from an abnormal state to a normal state can reflect the degree of action of the operating characteristic, therefore, the operational impact on the first performance index can be represented through the operating characteristic, therefore, the first optimization result can be based on the operating characteristic To further determine, obviously, in different application scenarios, for different performance indicators, the corresponding operating characteristics may also be different, which is not limited in this embodiment.

在一实施例中，第一优化结果包括针对第一性能指标的第一优化模型，还包括如下类型中的至少一种：针对第一性能指标的预测优化成功率；针对第一性能指标的实际优化成功率；与预测优化成功率对应的置信度。In one embodiment, the first optimization result includes the first optimization model for the first performance index, and also includes at least one of the following types: the predicted optimization success rate for the first performance index; the actual Optimization success rate; the confidence level corresponding to the predicted optimization success rate.

其中，针对第一性能指标，预测优化成功率为将相关特征参数输入到第一优化模型后而计算得到的预测优化成功率，其区别于实际优化成功率，通过比较两者之间的差异可以确定第一优化模型的精确程度；与预测优化成功率对应的置信度，可以表征达到预测优化成功率的可能性，即，对预测优化成功率进行区间估计，能够体现预测优化结果的稳定性，因此考虑置信度作为第一优化结果中的一项内容。Among them, for the first performance index, the predicted optimization success rate is the predicted optimization success rate calculated by inputting the relevant characteristic parameters into the first optimization model, which is different from the actual optimization success rate. By comparing the difference between the two, it can be Determine the accuracy of the first optimization model; the confidence corresponding to the predicted optimization success rate can represent the possibility of achieving the predicted optimization success rate, that is, the interval estimation of the predicted optimization success rate can reflect the stability of the predicted optimization results, Therefore, the confidence level is considered as an item in the first optimization result.

可以理解地是，第一优化结果还可以包括但不限于其他类型，比如提及到的作为样本的相关特征参数、优化过程中涉及到的其他参数等，这并未限制。It can be understood that the first optimization result may also include but not limited to other types, such as the relevant characteristic parameters mentioned as samples, other parameters involved in the optimization process, etc., which is not limited.

步骤S200:以所有第一优化结果和所有业务特征作为输入数据而构建联邦学习模型，并根据联邦学习模型确定第二优化模型。Step S200: Construct a federated learning model with all first optimization results and all business features as input data, and determine a second optimized model according to the federated learning model.

在一实施例中，由于所有第一优化结果和所有业务特征均是加密传输到第一服务器的，这确保了所有第一优化结果和所有业务特征的数据安全性，使得第一服务器根据这些加密数据而得到安全稳定的第二优化模型。In one embodiment, since all first optimization results and all business characteristics are encrypted and transmitted to the first server, this ensures the data security of all first optimization results and all business characteristics, so that the first server Data to obtain a safe and stable second optimization model.

在一实施例中，业务特征用于表征执行操作特征的环境信息，体现环境信息对于操作特征的影响，由于操作特征能够影响从异常状态恢复到正常状态的过程中的第一性能指标，因此，环境信息也能够间接地影响到第一性能指标的优化，所以在第一优化结果的基础上，进一步地以第一优化结果和业务特征作为输入数据而构建联邦学习模型，可以进一步地体现业务特征对于第一性能指标优化的影响，即能够基于联邦学习模型确定具有更高精确度的第二 优化模型。In one embodiment, the business feature is used to characterize the environmental information for executing the operating feature, reflecting the impact of the environmental information on the operating feature. Since the operating feature can affect the first performance index in the process of returning from an abnormal state to a normal state, therefore, Environmental information can also indirectly affect the optimization of the first performance index, so on the basis of the first optimization result, further use the first optimization result and business characteristics as input data to construct a federated learning model, which can further reflect business characteristics For the impact of the optimization of the first performance index, that is, the second optimization model with higher accuracy can be determined based on the federated learning model.

需要说明的是，不同的第二服务器所对应的业务特征互不影响，即，不同的第二服务器所对应的业务特征可能相同，也可能不相同，这是根据第二服务器的本地情况决定的。It should be noted that the service features corresponding to different second servers do not affect each other, that is, the service features corresponding to different second servers may be the same or may not be the same, which is determined according to the local conditions of the second server .

在一实施例中，环境信息可以是多元化的，即，所有影响于操作特征的其他因素均可以被看做是环境信息。In one embodiment, the environment information can be pluralized, that is, all other factors that affect the operating characteristics can be regarded as the environment information.

示例一example one

参照图4，以列表的形式示出了业务特征可能的不同类型的记录参数，可以理解地是，作为网络节点的基站，由于基站所提供的通信网络可能会影响到操作特征的执行，因此可以将基站的相关参数作为业务特征，用以表征相应的环境信息，例如，业务特征可以但不限于为基站所在经度、基站所在纬度、基站天线角度、基站海拔高度和基站正常运行时间长度等的一种或多种；或者，考虑网管系统自身的规模容量影响，当网管系统的规模较小时，可能会影响到操作特征的执行速度和流畅程度，即，业务特征还可以但不限于为网管或地区的规模、地区人口密度和小区与邻区的比例等的一种或多种。Referring to FIG. 4 , different types of recording parameters of service characteristics are shown in the form of a list. It can be understood that, as a base station of a network node, since the communication network provided by the base station may affect the execution of the operation characteristics, it can be The relevant parameters of the base station are used as service characteristics to represent the corresponding environmental information. For example, the service characteristics can be but not limited to a combination of the longitude of the base station, the latitude of the base station, the angle of the antenna of the base station, the altitude of the base station, and the length of normal operation of the base station. or, considering the scale and capacity of the network management system itself, when the scale of the network management system is small, it may affect the execution speed and smoothness of the operation characteristics, that is, the service characteristics can also be but not limited to network management or regional One or more of the size of the area, the population density of the area, and the ratio of the district to the neighborhood.

步骤S300:向各个第二服务器分别加密发送第二优化模型，以使各个第二服务器根据第二优化模型更新第一优化模型。Step S300: Encrypt and send the second optimization model to each second server, so that each second server updates the first optimization model according to the second optimization model.

在一实施例中，通过向各个第二服务器分别发送第二优化模型，使得各个第二服务器根据第二优化模型更新第一优化模型，从而能够在不泄露或转移数据的条件下进行性能指标优化，安全性更高，也能够提高第二服务器的性能指标优化成功率。In one embodiment, by sending the second optimization model to each second server, each second server updates the first optimization model according to the second optimization model, so that performance index optimization can be performed without leaking or transferring data , the security is higher, and the success rate of performance index optimization of the second server can also be improved.

在一实施例中，由于第二优化模型是经过联邦学习而得到的，因此第二优化模型相当于是基于不同的第二服务器的不同数据以整合得到的，能够符合不同的第二服务器的性能指标优化条件，结合前述实施例可知，不同的第二服务器仅为本地的目标数据服务，因此即使接收到第二优化模型，也不会对本地的第二服务器的性能指标优化产生干扰影响。In one embodiment, since the second optimization model is obtained through federated learning, the second optimization model is obtained based on different data from different second servers, and can meet the performance indicators of different second servers As for the optimization conditions, it can be seen from the foregoing embodiments that different second servers only serve local target data, so even if the second optimization model is received, it will not interfere with the performance index optimization of the local second server.

示例二Example two

本申请实施例提供的指标优化方法的流程可以是：The flow of the indicator optimization method provided in the embodiment of the present application may be:

假设存在两个第二服务器(记为A和B)，在确定A和B的情况下，A和B分别从相应的第一日志信息中提取出操作特征并确认得到第一优化结果，并分别将各自的操作特征和第一优化结果加密发送到对应的中间节点(第一中间节点对应于A，第二中间节点对应于B)，例如交换机、中转站等(可以是如图1中实施例所示的内网交换机)，同时将各自获取到的对应业务特征加密发送至相应的中间节点，然后第一服务器从第一中间节点和第二中间节点分别获取操作特征、第一优化结果和业务特征，并以所获取到的第一优化结果和业务特征作为输入数据而构建联邦学习模型，并根据联邦学习模型确定第二优化模型，最终第一服务器通过第一中间节点和第二中间节点分别向A和B对应返回第二优化模型，使得A和B分别根据第二优化模型更新第一优化模型，从而提高A和B的性能指标优化成功率。Assuming that there are two second servers (denoted as A and B), in the case of determining A and B, A and B respectively extract the operating characteristics from the corresponding first log information and confirm that the first optimization result is obtained, and respectively The respective operating characteristics and the first optimization result are encrypted and sent to the corresponding intermediate nodes (the first intermediate node corresponds to A, and the second intermediate node corresponds to B), such as switches, transfer stations, etc. (can be as shown in the embodiment in Figure 1 Intranet switch as shown), and at the same time, encrypt and send the corresponding service characteristics obtained respectively to the corresponding intermediate nodes, and then the first server obtains the operating characteristics, the first optimization result and the service characteristics from the first intermediate node and the second intermediate node respectively. features, and use the obtained first optimization results and business characteristics as input data to construct a federated learning model, and determine the second optimization model according to the federated learning model, and finally the first server passes the first intermediate node and the second intermediate node respectively The second optimization model is correspondingly returned to A and B, so that A and B respectively update the first optimization model according to the second optimization model, thereby improving the success rate of performance index optimization of A and B.

可以理解地是，步骤S100至S300是动态可循环执行的，即，各个第二服务器根据第二优化模型不断更新第一优化模型，并基于更新后的优化模型重新确定优化结果和业务特征，使得第一服务器持续地以所有第一优化结果和所有业务特征作为输入数据而构建联邦学习模型，并根据联邦学习模型确定不断优化的第二优化模型。It can be understood that steps S100 to S300 are dynamically and recursively executed, that is, each second server continuously updates the first optimization model according to the second optimization model, and re-determines the optimization result and business characteristics based on the updated optimization model, so that The first server continuously uses all first optimization results and all business features as input data to construct a federated learning model, and determines a continuously optimized second optimization model according to the federated learning model.

此外，在上述过程中，根据各个第二服务器在不同条件下所确定的优化结果的差异，可以进一步地对应确定业务特征的影响程度，从而决定是否仍以上一次的业务特征进行加密发 送，例如，第一优化结果中的预测优化成功率为79％，在经过第二优化模型更新第一优化模型之后，在原有的业务特征的基础上新增一个业务特征，在这种情况下，第二次获取到的更新后的第一优化结果的预测优化成功率为85％，则可以推断加入的新的业务特征对于提高预测优化成功率有着显著地影响，因此可以在以后的性能指标优化过程中，保留新添加的业务特征，类似地，若经过调整业务特征之后，第一优化结果的预测优化成功率反而相对降低，则可以选择剔除相应的业务特征。In addition, in the above process, according to the difference in the optimization results determined by each second server under different conditions, the degree of influence of the service characteristics can be further determined correspondingly, so as to determine whether to still encrypt and send the previous service characteristics, for example, The predicted optimization success rate in the first optimization result is 79%. After updating the first optimization model through the second optimization model, a new business feature is added on the basis of the original business feature. In this case, the second The predicted optimization success rate of the obtained updated first optimization result is 85%, it can be inferred that the new business features added have a significant impact on improving the predicted optimization success rate, so in the future performance index optimization process, The newly added business features are retained. Similarly, if the predicted optimization success rate of the first optimization result is relatively lower after the business features are adjusted, the corresponding business features can be selected to be eliminated.

如图5所示，图5是本申请另一个实施例提供的指标优化方法的流程图，该指标优化方法可以应用于如图1所示实施例中的网络拓扑中的第二服务器，第二服务器设置为多个，该方法包括但不限于步骤S400至S700。As shown in FIG. 5, FIG. 5 is a flow chart of an index optimization method provided by another embodiment of the present application. The index optimization method can be applied to the second server in the network topology in the embodiment shown in FIG. 1, and the second There are multiple servers, and the method includes but not limited to steps S400 to S700.

步骤S400:获取第一日志信息和业务特征，第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，业务特征用于表征执行操作特征的环境信息；Step S400: Obtain the first log information and business characteristics, the first log information carries the operation characteristics in the process of the first performance indicator recovering from the abnormal state to the normal state, and the business characteristics are used to represent the environmental information of the execution operation characteristics;

在一实施例中，通过获取第一日志信息和业务特征，可以确定与第一性能指标从异常状态恢复到正常状态的过程中的操作特征，以及该操作特征在执行条件下的环境信息，从而可以基于操作特征和环境信息在后续的步骤中实现构建联邦学习模型。In one embodiment, by acquiring the first log information and business characteristics, it is possible to determine the operating characteristics related to the first performance indicator in the process of returning from the abnormal state to the normal state, and the environment information of the operating characteristics under the execution conditions, so that Building a federated learning model can be implemented in subsequent steps based on operating characteristics and environmental information.

在一实施例中，操作特征包括如下类型中的至少一种：与第一性能指标对应的网元；与第一性能指标对应的对象；针对第一性能指标执行操作的名称；第一性能指标从异常状态恢复到正常状态前的数值；第一性能指标从异常状态恢复到正常状态后的数值；针对第一性能指标执行操作的结果。In an embodiment, the operation feature includes at least one of the following types: a network element corresponding to the first performance indicator; an object corresponding to the first performance indicator; a name of the operation performed on the first performance indicator; the first performance indicator The value before returning from the abnormal state to the normal state; the value of the first performance indicator after returning from the abnormal state to the normal state; the result of the operation performed on the first performance indicator.

需要说明的是，本实施例中的操作特征与上述如图3所示实施例的操作特征，具有相同的技术原理以及相同的技术效果，关于本实施例的技术原理以及技术效果，可以参照上述如图3所示实施例中的相关描述说明，为了避免内容重复冗余，此处不再赘述。It should be noted that the operational features in this embodiment have the same technical principles and the same technical effects as those of the above-mentioned embodiment shown in Figure 3. For the technical principles and technical effects of this embodiment, refer to the above Relevant descriptions in the embodiment shown in FIG. 3 illustrate that, in order to avoid content repetition and redundancy, details are not repeated here.

步骤S500:根据操作特征确定与第一性能指标对应的第一优化结果，第一优化结果包括针对第一性能指标的第一优化模型。Step S500: Determine a first optimization result corresponding to the first performance index according to the operating characteristics, the first optimization result includes a first optimization model for the first performance index.

在一实施例中，第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，由于该操作特征伴随着第一性能指标的状态变更过程，因此，可以认为该操作特征对于第一性能指标的优化有影响，因此，可以基于该操作特征来进一步地确认与第一性能指标对应的第一优化结果。In one embodiment, the first log information carries the operating characteristics of the first performance indicator in the process of recovering from the abnormal state to the normal state. Since the operating characteristics are accompanied by the state change process of the first performance indicator, it can be considered that the The operating characteristic has an influence on the optimization of the first performance index, therefore, the first optimization result corresponding to the first performance index can be further confirmed based on the operating characteristic.

在一实施例中，第一优化结果包括针对第一性能指标的第一优化模型，还包括如下类型中的至少一种：针对第一性能指标的预测优化成功率；针对第一性能指标的实际优化成功率；与预测优化成功率对应的置信度。In one embodiment, the first optimization result includes the first optimization model for the first performance index, and also includes at least one of the following types: the predicted optimization success rate for the first performance index; the actual Optimization success rate; the confidence level corresponding to the predicted optimization success rate.

可以理解地是，第一优化结果还可以包括但不限于其他类型，比如作为样本的相关特征参数、优化过程中涉及到的其他参数等，这并未限制。It can be understood that the first optimization result may also include but not limited to other types, such as relevant characteristic parameters as samples, other parameters involved in the optimization process, etc., which is not limited.

在图6的示例中，在操作特征为多个，第一日志信息还包括指标影响特征，指标影响特征用于表征多个操作特征之间的相关性的情况下，步骤S500包括但不限于步骤S510至S520。In the example of FIG. 6 , when there are multiple operating features, and the first log information also includes index impact features, and the index impact features are used to characterize the correlation between multiple operating features, step S500 includes but is not limited to the step S510 to S520.

步骤S510:根据指标影响特征从各个操作特征中确定具有相关性的多个操作特征。Step S510: Determine a plurality of relevant operating features from various operating features according to the index impact feature.

步骤S520:根据具有相关性的多个操作特征确定与第一性能指标对应的第一优化结果。Step S520: Determine a first optimization result corresponding to a first performance index according to a plurality of correlated operating characteristics.

在一实施例中，当操作特征为多个，则意味着可能出现同时执行不同的操作特征，在这种情况下，若不同操作特征之间的配合效果更好，则有利于对第一性能指标进行更准确地优化，因此，引入了指标影响特征这一概念，可以通过指标影响特征确定多个操作特征之间的 相关性，从而将具有相关性的多个操作特征提取出来配合执行，可以免除不相关的操作特征的影响，使得操作特征的执行效果将会更加良好，其中，指标影响特征可以通过采样统计而确定，可以以相关系数或等级的形式进行表示，在判断时可以设置阈值来选出相关的操作特征，例如，针对图3所示实施例提供的操作特征，若指标影响特征表征网元出现次数与操作对象、操作结果之间的相关系数，同时预设一个相关系数阈值，当相关系数不小于相关系数阈值，则可以判定网元、操作对象和操作结果之间的相关程度较高，可确定网元、操作对象和操作结果为一组相关操作特征，进而根据具有相关性的网元、操作对象和操作结果确定与第一性能指标对应的第一优化结果。In one embodiment, when there are multiple operating features, it means that different operating features may be executed at the same time. In this case, if the cooperation effect between different operating features is better, it is beneficial to the first performance Indexes are optimized more accurately. Therefore, the concept of index influence features is introduced, and the correlation between multiple operational features can be determined through the index influence features, so that multiple operational features with correlation can be extracted to cooperate with execution. Exempting the influence of irrelevant operating characteristics will make the execution effect of operating characteristics better. Among them, the characteristics of index influence can be determined through sampling statistics, and can be expressed in the form of correlation coefficient or level. Thresholds can be set when judging. Select relevant operating features, for example, for the operating features provided in the embodiment shown in Figure 3, if the index impact feature represents the correlation coefficient between the number of network element occurrences, the operation object, and the operation result, and a correlation coefficient threshold is preset at the same time, When the correlation coefficient is not less than the correlation coefficient threshold, it can be determined that the correlation between the network element, the operation object and the operation result is high, and it can be determined that the network element, the operation object and the operation result are a set of related operation characteristics, and then according to the correlation The first optimization result corresponding to the first performance index is determined by the network element, the operation object and the operation result.

在一实施例中，针对所选择的多个操作特征进行指标优化预测时，通过将网元、操作对象和操作结果作为输入数据提供给第一优化模型，第一优化模型则会相应记录与输入信息相关的内容，并输出针对第一性能指标的预测优化成功率，基于以上信息则可以确定第一优化结果的全部内容。In one embodiment, when performing index optimization prediction for multiple selected operating characteristics, by providing network elements, operation objects and operation results as input data to the first optimization model, the first optimization model will record and input the corresponding Information-related content, and output the predicted optimization success rate for the first performance index, based on the above information, the entire content of the first optimization result can be determined.

步骤S600:向第一服务器加密传输第一优化结果和业务特征，以使第一服务器根据联邦学习模型确定第二优化模型，联邦学习模型为由第一服务器以多个第二服务器分别发送的第一优化结果和业务特征作为输入数据而构建；Step S600: Encrypted transmission of the first optimization result and business characteristics to the first server, so that the first server determines the second optimization model according to the federated learning model, the federated learning model is the first optimization result sent by the first server and a plurality of second servers respectively. - optimization results and business characteristics are constructed as input data;

在一实施例中，通过向第一服务器加密传输第一优化结果和业务特征，确保了所有第一优化结果和所有业务特征的数据安全性，使得第一服务器所确定的第二优化模型具有数据安全性，同时由于将业务特征与第一优化结果实现配合传输，即，以第一优化结果和业务特征作为输入数据而构建联邦学习模型，可以进一步地体现业务特征对于第一性能指标优化的影响，即能够基于联邦学习模型确定具有更高精确度的第二优化模型。In one embodiment, the data security of all first optimization results and all business features is ensured by encrypting and transmitting the first optimization results and business features to the first server, so that the second optimization model determined by the first server has data Security, at the same time, because the business characteristics and the first optimization result are transmitted together, that is, the federated learning model is constructed with the first optimization result and business characteristics as input data, which can further reflect the impact of business characteristics on the optimization of the first performance index , that is, the second optimization model with higher accuracy can be determined based on the federated learning model.

在图7的示例中，步骤S600包括但不限于步骤S610至步骤S620。In the example of FIG. 7, step S600 includes but not limited to step S610 to step S620.

步骤S610:根据第一优化结果和业务特征确定与第一优化结果和业务特征对应的第一梯度。Step S610: Determine the first gradient corresponding to the first optimization result and the business feature according to the first optimization result and the business feature.

步骤S620:向第一服务器加密传输第一梯度。Step S620: encrypt and transmit the first gradient to the first server.

在一实施例中，由于梯度作为一个向量参数，其既表征大小也表征方向，因此梯度作为数据参数具有可靠的稳定性，因此，采用第一梯度来承载第一优化结果和业务特征，可以简化传输难度，当其经过加密传输至第一服务器之后，第一服务器可以还原第一梯度而得到原来的第一优化结果和业务特征，较为方便可靠。In an embodiment, since the gradient is used as a vector parameter, which represents both the magnitude and the direction, the gradient has reliable stability as a data parameter. Therefore, using the first gradient to carry the first optimization result and business characteristics can simplify Difficulty in transmission. After encrypted transmission to the first server, the first server can restore the first gradient to obtain the original first optimization result and business characteristics, which is more convenient and reliable.

需要说明的是，上述各实施例中的加密所对应的具体手段并不限定，例如差异隐私化、秘密共享等，可以根据实际应用场景自行设置。It should be noted that the specific means corresponding to the encryption in the above embodiments are not limited, such as differential privacy, secret sharing, etc., which can be set according to actual application scenarios.

步骤S700：接收由第一服务器加密发送的第二优化模型，并根据第二优化模型更新第一优化模型。Step S700: Receive the second optimization model encrypted and sent by the first server, and update the first optimization model according to the second optimization model.

在一实施例中，通过接收第二优化模型，从而可以根据第二优化模型更新第一优化模型，从而能够在不泄露或转移数据的条件下进行性能指标优化，安全性更高，也能够提高第二服务器的性能指标优化成功率。In one embodiment, by receiving the second optimization model, the first optimization model can be updated according to the second optimization model, so that performance index optimization can be performed without leaking or transferring data, which has higher security and can also improve The performance index optimization success rate of the second server.

另外，该指标优化方法可以应用于如图1所示实施例中的网络拓扑中的第二服务器和第一服务器，该方法包括但不限于步骤S800至S1200。In addition, the index optimization method can be applied to the second server and the first server in the network topology in the embodiment shown in FIG. 1 , and the method includes but is not limited to steps S800 to S1200.

步骤S800:第二服务器获取第一日志信息和业务特征，第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，业务特征用于表征执行操作特征的环境 信息。Step S800: the second server obtains the first log information and business characteristics, the first log information carries the operation characteristics of the first performance index in the process of recovering from the abnormal state to the normal state, and the business characteristics are used to represent the environmental information of the execution operation characteristics .

步骤S900:第二服务器根据操作特征确定与第一性能指标对应的第一优化结果，并向第一服务器加密传输第一优化结果和业务特征，第一优化结果包括针对第一性能指标的第一优化模型。Step S900: The second server determines the first optimization result corresponding to the first performance index according to the operating characteristics, and encrypts and transmits the first optimization result and business characteristics to the first server, the first optimization result includes the first optimization result for the first performance index Optimize the model.

步骤S1000:第一服务器以所有第二服务器的第一优化结果和业务特征作为输入数据而构建联邦学习模型，并根据联邦学习模型确定第二优化模型。Step S1000: the first server constructs a federated learning model with the first optimization results and service characteristics of all second servers as input data, and determines a second optimized model according to the federated learning model.

步骤S1100:向第二服务器加密发送第二优化模型。Step S1100: encrypt and send the second optimization model to the second server.

步骤S1200:第二服务器根据第二优化模型更新第一优化模型。Step S1200: the second server updates the first optimization model according to the second optimization model.

需要说明的是，本实施例中的步骤S800至S1200与上述如图2所示实施例的步骤S100至S300以及如图5所示实施例的步骤S400至S700，具有相同的技术原理以及相同的技术效果，不同实施例之间的区别在于执行主体不同，其中，上述如图2所示实施例的执行主体为第一服务器，上述如图5所示实施例的执行主体为第二服务器，而本实施例的执行主体为第一服务器和第二服务器。关于本实施例的技术原理以及技术效果，可以参照上述如图2和图5所示实施例中的相关描述说明，为了避免内容重复冗余，此处不再赘述。It should be noted that steps S800 to S1200 in this embodiment have the same technical principle and the same Technical effect, the difference between different embodiments is that the execution subject is different, wherein, the execution subject of the embodiment shown in Figure 2 is the first server, the execution subject of the embodiment shown in Figure 5 is the second server, and The execution subjects of this embodiment are the first server and the second server. Regarding the technical principles and technical effects of this embodiment, reference may be made to relevant descriptions in the above-mentioned embodiments shown in FIG. 2 and FIG. 5 .

另外，参照图8，本申请的一个实施例还提供了一种服务器，该服务器包括：第一存储器、第一处理器及存储在第一存储器上并可在第一处理器上运行的计算机程序。In addition, referring to FIG. 8 , an embodiment of the present application also provides a server, the server includes: a first memory, a first processor, and a computer program stored on the first memory and operable on the first processor .

第一处理器和第一存储器可以通过第一总线或者其他方式连接。The first processor and the first memory may be connected through a first bus or in other ways.

需要说明的是，本实施例中的服务器，可以应用为例如图1所示实施例中的第一服务器，本实施例中的服务器能够构成例如图1所示实施例中的网络拓扑的一部分，这些实施例均属于相同的发明构思，因此这些实施例具有相同的实现原理以及技术效果，此处不再详述。It should be noted that the server in this embodiment can be applied as the first server in the embodiment shown in FIG. 1, and the server in this embodiment can constitute a part of the network topology in the embodiment shown in FIG. 1, for example. These embodiments all belong to the same inventive concept, so these embodiments have the same implementation principle and technical effect, and will not be described in detail here.

实现上述实施例的指标优化方法所需的非暂态软件程序以及指令存储在第一存储器中，当被第一处理器执行时，执行上述各实施例的指标优化方法，例如，执行以上描述的图2中的方法步骤S100至S300。The non-transitory software programs and instructions required to implement the index optimization method of the above-mentioned embodiments are stored in the first memory, and when executed by the first processor, the index optimization methods of the above-mentioned embodiments are executed, for example, the above-described Method steps S100 to S300 in FIG. 2 .

另外，参照图9，本申请的一个实施例还提供了一种服务器，该服务器包括：第二存储器、第二处理器及存储在第二存储器上并可在第二处理器上运行的计算机程序。In addition, referring to FIG. 9 , an embodiment of the present application also provides a server, the server includes: a second memory, a second processor, and a computer program stored on the second memory and operable on the second processor .

第二处理器和第二存储器可以通过第二总线或者其他方式连接。The second processor and the second memory may be connected through a second bus or in other ways.

需要说明的是，本实施例中的服务器，可以应用为例如图1所示实施例中的第二服务器，本实施例中的服务器能够构成例如图1所示实施例中的网络拓扑的一部分，这些实施例均属于相同的发明构思，因此这些实施例具有相同的实现原理以及技术效果，此处不再详述。It should be noted that the server in this embodiment can be applied as the second server in the embodiment shown in FIG. 1, and the server in this embodiment can constitute a part of the network topology in the embodiment shown in FIG. 1, for example. These embodiments all belong to the same inventive concept, so these embodiments have the same implementation principle and technical effect, and will not be described in detail here.

实现上述实施例的指标优化方法所需的非暂态软件程序以及指令存储在第二存储器中，当被第二处理器执行时，执行上述各实施例的指标优化方法，例如，执行以上描述的图5中的方法步骤S400至S700、图6中的方法步骤S510至S520或图7中的方法步骤S610至S620。The non-transitory software programs and instructions required to realize the index optimization methods of the above-mentioned embodiments are stored in the second memory, and when executed by the second processor, the index optimization methods of the above-mentioned embodiments are executed, for example, the above-described Method steps S400 to S700 in FIG. 5 , method steps S510 to S520 in FIG. 6 , or method steps S610 to S620 in FIG. 7 .

另外，本申请的一个实施例还提供了一种网络节点，该网络节点包括：第三存储器、第三处理器及存储在第三存储器上并可在第三处理器上运行的计算机程序。In addition, an embodiment of the present application further provides a network node, where the network node includes: a third memory, a third processor, and a computer program stored in the third memory and operable on the third processor.

第三处理器和第三存储器可以通过第三总线或者其他方式连接。The third processor and the third memory may be connected through a third bus or in other ways.

需要说明的是，本实施例中的网络节点，可以应用为例如图1所示实施例中的第一服务器和第二服务器，本实施例中的网络节点能够构成例如图1所示实施例中的网络拓扑的一部分，这些实施例均属于相同的发明构思，因此这些实施例具有相同的实现原理以及技术效果，此处不再详述。It should be noted that the network nodes in this embodiment can be applied to, for example, the first server and the second server in the embodiment shown in Figure 1, and the network nodes in this embodiment can constitute These embodiments all belong to the same inventive concept, so these embodiments have the same implementation principle and technical effect, and will not be described in detail here.

实现上述实施例的指标优化方法所需的非暂态软件程序以及指令存储在第三存储器中，当被第三处理器执行时，执行上述各实施例的指标优化方法，例如，执行以上描述的方法步骤S800至S1200。The non-transitory software programs and instructions required to realize the index optimization method of the above-mentioned embodiments are stored in the third memory, and when executed by the third processor, the index optimization methods of the above-mentioned embodiments are executed, for example, the above-described Method steps S800 to S1200.

以上所描述的装置实施例仅仅是示意性的，其中作为分离部件说明的单元可以是或者也可以不是物理上分开的，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

此外，本申请的一个实施例还提供了一种计算机可读存储介质，该计算机可读存储介质存储有计算机可执行指令，该计算机可执行指令被一个第一处理器、第二处理器、第三处理器或控制器执行，例如，被上述设备实施例中的一个第一处理器、第二处理器或第三处理器执行，可使得上述第一处理器、第二处理器或第三处理器执行上述实施例中的指标优化方法，例如，执行以上描述的图2中的方法步骤S100至S300，或者，执行以上描述的图5中的方法步骤S400至S700、图6中的方法步骤S510至S520或图7中的方法步骤S610至S620，或者，执行以上描述的方法步骤S800至S1200。In addition, an embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a first processor, a second processor, and a second processor. Execution by three processors or controllers, for example, being executed by a first processor, a second processor or a third processor in the above-mentioned device embodiment, may cause the above-mentioned first processor, second processor or third processing The controller executes the index optimization method in the above embodiment, for example, executes the method steps S100 to S300 in FIG. 2 described above, or executes the method steps S400 to S700 in FIG. 5 and the method step S510 in FIG. 6 described above to S520 or the method steps S610 to S620 in FIG. 7 , or, perform the method steps S800 to S1200 described above.

本申请实施例包括：应用于第一服务器的指标优化方法，包括获取由各个第二服务器分别加密发送的与第一性能指标对应的第一优化结果和业务特征，其中，第一优化结果为由第二服务器的第一日志信息确定，第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，业务特征用于表征执行操作特征的环境信息，第一优化结果包括针对第一性能指标的第一优化模型；以所有第一优化结果和所有业务特征作为输入数据而构建联邦学习模型，并根据联邦学习模型确定第二优化模型；向各个第二服务器分别加密发送第二优化模型，以使各个第二服务器根据第二优化模型更新第一优化模型。根据本申请实施例提供的方案，通过获取由各个第二服务器发送的第一优化结果和业务特征，使得在确定的第一优化结果的基础上结合业务特征而构建联邦学习模型，从而基于联邦学习模型可以得到相比第一优化模型具有更高精确度的第二优化模型，相比于相关技术，能够在不泄露或转移数据的条件下进行性能指标优化，安全性更高，并通过向各个第二服务器分别发送第二优化模型，使得各个第二服务器根据第二优化模型更新第一优化模型，从而能够提高第二服务器的性能指标优化成功率。The embodiment of the present application includes: an index optimization method applied to the first server, including obtaining the first optimization result and service characteristics corresponding to the first performance index encrypted and sent by each second server, wherein the first optimization result is obtained by The first log information of the second server is determined. The first log information carries the operating characteristics of the first performance index in the process of returning from the abnormal state to the normal state. The business characteristics are used to represent the environmental information of the execution operation characteristics. The first optimization result Including the first optimization model for the first performance index; using all the first optimization results and all business characteristics as input data to construct a federated learning model, and determining the second optimized model according to the federated learning model; sending encrypted data to each second server The second optimization model, so that each second server updates the first optimization model according to the second optimization model. According to the solution provided by the embodiment of the present application, by obtaining the first optimization results and business characteristics sent by each second server, the federated learning model is constructed on the basis of the determined first optimization results combined with the business characteristics, so that based on federated learning The model can obtain a second optimization model with higher accuracy than the first optimization model. Compared with related technologies, it can optimize performance indicators without leaking or transferring data, and has higher security. The second servers respectively send the second optimization models, so that each second server updates the first optimization model according to the second optimization models, thereby improving the success rate of performance index optimization of the second servers.

本领域普通技术人员可以理解，上文中所公开方法中的全部或某些步骤、系统可以被实施为软件、固件、硬件及其适当的组合。某些物理组件或所有物理组件可以被实施为由处理器，如中央处理器、数字信号处理器或微处理器执行的软件，或者被实施为硬件，或者被实施为集成电路，如专用集成电路。这样的软件可以分布在计算机可读介质上，计算机可读介质可以包括计算机存储介质(或非暂时性介质)和通信介质(或暂时性介质)。如本领域普通技术人员公知的，术语计算机存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储装置、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。此外，本领域普通技术人员公知的是，通信介质通常包含计算机可读指令、数据结构、程序模块或者诸如载波或其他传输机制之类的调制数据信号中的其他数据，并且可包括任何信息递送介质。Those skilled in the art can understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware and an appropriate combination thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit . Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those of ordinary skill in the art, the term computer storage media includes both volatile and nonvolatile media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. permanent, removable and non-removable media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, tape, magnetic disk storage or other magnetic storage devices, or can Any other medium used to store desired information and which can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media .

以上是对本申请的若干实施方式进行的具体说明，但本申请并不局限于上述实施方式， 熟悉本领域的技术人员在不违背本申请精神的前提下还可作出种种的等同变形或替换，这些等同的变形或替换均包含在本申请权利要求所限定的范围内。The above is a specific description of several implementations of the present application, but the application is not limited to the above-mentioned implementations, and those skilled in the art can make various equivalent deformations or replacements without violating the spirit of the application. Equivalent modifications or replacements are all within the scope defined by the claims of the present application.

Claims (11)

Translated fromChinese

一种指标优化方法，应用于第一服务器，所述方法包括：An index optimization method applied to a first server, the method comprising:获取由各个第二服务器分别加密发送的与第一性能指标对应的第一优化结果和业务特征，其中，所述第一优化结果为由所述第二服务器的第一日志信息确定，所述第一日志信息携带有所述第一性能指标从异常状态恢复到正常状态的过程中的操作特征，所述业务特征用于表征执行所述操作特征的环境信息，所述第一优化结果包括针对所述第一性能指标的第一优化模型；Obtain the first optimization results and service characteristics corresponding to the first performance indicators encrypted and sent by each second server, wherein the first optimization results are determined by the first log information of the second server, and the first A log information carries operating characteristics during the process of restoring the first performance indicator from an abnormal state to a normal state, the business characteristics are used to characterize the environment information for executing the operating characteristics, and the first optimization result includes The first optimization model of the first performance index;以所有所述第一优化结果和所有所述业务特征作为输入数据而构建联邦学习模型，并根据所述联邦学习模型确定第二优化模型；Constructing a federated learning model with all of the first optimization results and all of the business characteristics as input data, and determining a second optimization model according to the federated learning model;向各个所述第二服务器分别加密发送所述第二优化模型，以使各个所述第二服务器根据所述第二优化模型更新所述第一优化模型。The second optimization model is encrypted and sent to each of the second servers, so that each of the second servers updates the first optimization model according to the second optimization model.根据权利要求1所述的指标优化方法，其中，所述第一优化结果还包括如下类型中的至少一种：The index optimization method according to claim 1, wherein the first optimization result further includes at least one of the following types:针对所述第一性能指标的预测优化成功率；A prediction optimization success rate for the first performance index;针对所述第一性能指标的实际优化成功率；The actual optimization success rate for the first performance indicator;与所述预测优化成功率对应的置信度。Confidence corresponding to the predicted optimization success rate.根据权利要求1或2所述的指标优化方法，其中，所述操作特征包括如下类型中的至少一种：The index optimization method according to claim 1 or 2, wherein the operating characteristics include at least one of the following types:与所述第一性能指标对应的网元；A network element corresponding to the first performance indicator;与所述第一性能指标对应的对象；an object corresponding to the first performance indicator;针对所述第一性能指标执行操作的名称；the name of the operation performed on the first performance indicator;所述第一性能指标从异常状态恢复到正常状态前的数值；The value before the first performance index recovers from the abnormal state to the normal state;所述第一性能指标从异常状态恢复到正常状态后的数值；The value of the first performance index after returning from an abnormal state to a normal state;针对所述第一性能指标执行操作的结果。A result of performing an operation on the first performance indicator.一种指标优化方法，应用于第二服务器，所述第二服务器设置有多个，所述方法包括：An index optimization method, applied to a second server, where multiple second servers are provided, the method includes:获取第一日志信息和业务特征，所述第一日志信息携带有第一性能指标从异常状态恢复到正常状态的过程中的操作特征，所述业务特征用于表征执行所述操作特征的环境信息；Acquiring first log information and service characteristics, the first log information carrying the operation characteristics of the first performance indicator in the process of recovering from the abnormal state to the normal state, the business characteristics are used to represent the environment information for executing the operation characteristics ;根据所述操作特征确定与所述第一性能指标对应的第一优化结果，所述第一优化结果包括针对所述第一性能指标的第一优化模型；determining a first optimization result corresponding to the first performance index according to the operating characteristics, the first optimization result including a first optimization model for the first performance index;向第一服务器加密传输所述第一优化结果和所述业务特征，以使所述第一服务器根据联邦学习模型确定第二优化模型，所述联邦学习模型为由所述第一服务器以多个所述第二服务器分别发送的所述第一优化结果和所述业务特征作为输入数据而构建；encrypting and transmitting the first optimization result and the business characteristics to the first server, so that the first server determines a second optimization model according to the federated learning model, the federated learning model is generated by the first server with multiple The first optimization result and the service characteristics respectively sent by the second server are constructed as input data;接收由所述第一服务器加密发送的所述第二优化模型，并根据所述第二优化模型更新所述第一优化模型。receiving the second optimization model encrypted and sent by the first server, and updating the first optimization model according to the second optimization model.根据权利要求4所述的一种指标优化方法，其中，所述操作特征为多个，所述第一日志信息还包括指标影响特征，所述指标影响特征用于表征多个操作特征之间的相关性；The index optimization method according to claim 4, wherein there are multiple operating features, and the first log information further includes index impact features, and the index impact features are used to characterize the relationship between multiple operating features Correlation;所述根据所述操作特征确定与所述第一性能指标对应的第一优化结果，包括：The determining the first optimization result corresponding to the first performance index according to the operating characteristics includes:根据所述指标影响特征从各个所述操作特征中确定具有相关性的多个所述操作特征；determining a plurality of relevant operational characteristics from each of the operational characteristics according to the index impact characteristics;根据具有相关性的多个所述操作特征确定与所述第一性能指标对应的第一优化结果。A first optimization result corresponding to the first performance index is determined according to a plurality of the operating characteristics with correlation.根据权利要求4所述的一种指标优化方法，其中，所述向第一服务器加密传输所述第一优化结果和所述业务特征，包括：The index optimization method according to claim 4, wherein said encrypted transmission of said first optimization result and said service characteristics to the first server comprises:根据所述第一优化结果和所述业务特征确定与所述第一优化结果和所述业务特征对应的第一梯度；determining a first gradient corresponding to the first optimization result and the service feature according to the first optimization result and the service feature;向第一服务器加密传输所述第一梯度。Encrypted transmission of the first gradient to the first server.根据权利要求4所述的指标优化方法，其中，所述第一优化结果还包括如下类型中的至少一种：The index optimization method according to claim 4, wherein the first optimization result further includes at least one of the following types:针对所述第一性能指标的预测优化成功率；For the predicted optimization success rate of the first performance index;针对所述第一性能指标的实际优化成功率；The actual optimization success rate for the first performance index;与所述预测优化成功率对应的置信度。Confidence corresponding to the predicted optimization success rate.根据权利要求4或7所述的指标优化方法，其中，所述操作特征包括如下类型中的至少一种：The index optimization method according to claim 4 or 7, wherein the operating characteristics include at least one of the following types:与所述第一性能指标对应的网元；A network element corresponding to the first performance indicator;与所述第一性能指标对应的对象；an object corresponding to the first performance indicator;针对所述第一性能指标执行操作的名称；the name of the operation performed on the first performance indicator;所述第一性能指标从异常状态恢复到正常状态前的数值；The value before the first performance index recovers from the abnormal state to the normal state;所述第一性能指标从异常状态恢复到正常状态后的数值；The value of the first performance index after returning from an abnormal state to a normal state;针对所述第一性能指标执行操作的结果。A result of performing an operation on the first performance indicator.一种服务器，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如权利要求1至3中任意一项所述的指标优化方法。A server, comprising: a memory, a processor, and a computer program stored on the memory and operable on the processor, when the processor executes the computer program, the method described in any one of claims 1 to 3 is realized Index optimization method.一种服务器，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如权利要求4至8中任意一项所述的指标优化方法。A server, comprising: a memory, a processor, and a computer program stored on the memory and operable on the processor, when the processor executes the computer program, the method described in any one of claims 4 to 8 is realized Index optimization method.一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行权利要求1至3中任意一项所述的指标优化方法，或者，执行权利要求4至8中任意一项所述的指标优化方法。A computer-readable storage medium, storing computer-executable instructions, the computer-executable instructions are used to execute the index optimization method described in any one of claims 1 to 3, or to execute any one of claims 4 to 8 One of the described index optimization methods.

Images