Failure prediction method and device for financial service, storage medium and electronic equipment

Abstract

The invention discloses a fault prediction method and device for financial service, a storage medium and electronic equipment. The method relates to the field of artificial intelligence, and comprises the following steps: obtaining service data of a target financial service in a business platform in N time ranges, the service data at least comprising a service log and service performance data, and N being a positive integer greater than 1; performing feature extraction processing on the service data in each time range to obtain data features of each piece of service data; the data features of the N service data are processed through a fault prediction model, a fault prediction result is obtained, and the fault prediction result represents the probability that the target financial service has different fault types; target information is determined based on the fault prediction result, and the target information represents whether the target financial service breaks down or not and the fault type of the target financial service under the condition that the target financial service breaks down. Through the method and the device, the problem of low accuracy of fault prediction of financial services in related technologies is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more specifically, to a method, apparatus, storage medium, and electronic device for predicting faults in financial services. Background Technology

[0002] In the context of the rapid development of the financial services industry, various online transaction and payment services, along with their back-end support systems, constitute a complex IT (Information Technology) ecosystem. These systems operate 24 / 7, handling the financial activities of a massive number of users. If financial service malfunctions or impending malfunctions are not detected in a timely manner, it will not only impact user experience but also cause significant losses to financial institutions. Currently, related technologies rely on monitoring data from financial services and threshold comparisons to determine whether malfunctions have occurred, resulting in low accuracy in malfunction prediction.

[0003] There is currently no effective solution to the aforementioned problems in the relevant technologies. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, storage medium, and electronic device for predicting failures in financial services, in order to solve the problem of low accuracy in predicting failures in financial services in related technologies.

[0005] To achieve the above objectives, according to one aspect of this application, a method for predicting failures in financial services is provided. The method includes: acquiring service data of a target financial service within N time ranges from a business platform, wherein the service data includes at least: service logs and service performance data, and N is a positive integer greater than 1; performing feature extraction processing on the service data within each time range to obtain data features for each service data; processing the data features of the N service data using a failure prediction model to obtain failure prediction results, wherein the failure prediction results characterize the probability of different failure types occurring in the target financial service; and determining target information based on the failure prediction results, wherein the target information characterizes whether the target financial service has failed, and the type of failure in the event of a failure.

[0006] Furthermore, the fault prediction method for financial services also includes: for service data within each time range, extracting features from service logs using a word embedding model to obtain a first vector; dividing the time range into M time windows based on a preset sliding window, and calculating the target features of service performance data in each time window, wherein the target features include at least one of the following types of features: mean, variance, skewness, kurtosis, first difference, second difference, where M is a positive integer greater than 1; processing the target features of service performance data in the M time windows using a time-series feature extractor to obtain a second vector; and determining the data features of service data based on the first vector and the second vector.

[0007] Furthermore, the fault prediction method for financial services also includes: processing the system topology graph through a graph neural network to obtain a third vector; and determining the data characteristics of the service data based on the first vector, the second vector, and the third vector.

[0008] Furthermore, the fault prediction method for financial services also includes: acquiring a target training sample set, wherein the training samples in the target training sample set are sample service data of the sample financial services, and the true labels of the training samples represent the fault types of the sample financial services; processing the target training sample set through an initial fault prediction model to obtain the sample prediction results output by the output layer of the initial fault prediction model and the intermediate features output by the first processing layer; processing the fault description information of the training samples through a fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model, wherein the fault analysis model is a large language model, and the fault analysis model is used to determine the cause of the fault and the scope of the fault impact based on the fault description information; determining the loss function value based on the difference information between the sample prediction results and the true labels, and the difference information between the fault features and the intermediate features, and optimizing the model parameters of the initial fault prediction model based on the loss function value until the loss function value is less than a preset value, thereby obtaining the fault prediction model.

[0009] Furthermore, the failure prediction method for financial services also includes: determining a first loss value based on a first sub-loss function, sample prediction results, and true labels; determining a second loss value based on a second sub-loss function, failure features, and intermediate features; and determining a loss function value based on the first loss value, the second loss value, a first coefficient associated with the first sub-loss function, and a second coefficient associated with the second sub-loss function.

[0010] Furthermore, the fault prediction method for financial services also includes: determining whether there is a target probability value in the fault prediction result, wherein the target probability value is a probability value that is greater than a preset probability; if there is a target probability value in the fault prediction result, determining that the target information represents a fault in the target financial service, and determining the fault type corresponding to the highest target probability value as the fault type of the target financial service; if there is no target probability value in the fault prediction result, determining that the target information represents a fault in the target financial service.

[0011] Furthermore, the fault prediction method for financial services also includes: acquiring system resource information of the target financial service, and composing fault information based on the highest target probability value, the fault type of the target financial service, and the system resource information of the target financial service; processing the fault information through a fault handling model to obtain target operation information, wherein the target operation information includes operation information for repairing the fault of the target financial service, and the target financial service is trained based on a preset state space, action space, and reward function, the state space is constructed based on historical fault information, and the action space is constructed based on the operation information of multiple preset operations.

[0012] To achieve the above objectives, according to another aspect of this application, a fault prediction device for financial services is provided. The device includes: a first acquisition module, configured to acquire service data of a target financial service within N time ranges from a business platform, wherein the service data includes at least: service logs and service performance data, and N is a positive integer greater than 1; a first processing module, configured to perform feature extraction processing on the service data within each time range to obtain data features for each service data; a second processing module, configured to process the data features of the N service data using a fault prediction model to obtain fault prediction results, wherein the fault prediction results characterize the probability of different fault types occurring in the target financial service; and a first determination module, configured to determine target information based on the fault prediction results, wherein the target information characterizes whether the target financial service has experienced a fault, and the type of fault in the event of a fault.

[0013] Furthermore, the first processing module also includes: a first extraction submodule, used to extract features from service logs using a word embedding model for service data within each time range to obtain a first vector; a calculation submodule, used to divide the time range into M time windows based on a preset sliding window, and calculate the target features of service performance data in each time window, wherein the target features include at least one of the following types of features: mean, variance, skewness, kurtosis, first difference, second difference, where M is a positive integer greater than 1; a second extraction submodule, used to process the target features of service performance data in the M time windows using a time-series feature extractor to obtain a second vector; and a first determination submodule, used to determine the data features of service data based on the first vector and the second vector.

[0014] Furthermore, the first determining submodule also includes: a processing unit, used to process the system topology graph through a graph neural network to obtain a third vector; and a determining unit, used to determine the data features of the service data based on the first vector, the second vector, and the third vector.

[0015] Furthermore, the fault prediction device for financial services also includes: a second acquisition module for acquiring a target training sample set, wherein the training samples in the target training sample set are sample service data of the sample financial services, and the true labels of the training samples represent the fault types of the sample financial services; a third processing module for processing the target training sample set through an initial fault prediction model to obtain the sample prediction results output by the output layer of the initial fault prediction model and the intermediate features output by the first processing layer; a fourth processing module for processing the fault description information of the training samples through a fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model, wherein the fault analysis model is a large language model, and the fault analysis model is used to determine the cause of the fault and the scope of its impact based on the fault description information; and a second determination module for determining the loss function value based on the difference information between the sample prediction results and the true labels, and the difference information between the fault features and the intermediate features, and optimizing the model parameters of the initial fault prediction model based on the loss function value until the loss function value is less than a preset value, thereby obtaining the fault prediction model.

[0016] Furthermore, the second determining module also includes: a second determining submodule, used to determine a first loss value based on the first sub-loss function, the sample prediction result, and the true label; a third determining submodule, used to determine a second loss value based on the second sub-loss function, fault features, and intermediate features; and a fourth determining submodule, used to determine a loss function value based on the first loss value, the second loss value, a first coefficient associated with the first sub-loss function, and a second coefficient associated with the second sub-loss function.

[0017] Furthermore, the first determining module also includes: a judging submodule, used to judge whether there is a target probability value in the fault prediction result, wherein the target probability value is a probability value greater than a preset probability; a fifth determining submodule, used to determine that the target information represents a fault in the target financial service when there is a target probability value in the fault prediction result, and to determine the fault type corresponding to the highest target probability value as the fault type of the target financial service; and a sixth determining submodule, used to determine that the target information represents no fault in the target financial service when there is no target probability value in the fault prediction result.

[0018] Furthermore, the fault prediction device for financial services also includes: a third acquisition module, used to acquire system resource information of the target financial service, and to compose fault information based on the highest target probability value, the fault type of the target financial service, and the system resource information of the target financial service; and a fifth processing module, used to process the fault information through a fault processing model to obtain target operation information, wherein the target operation information includes operation information for repairing the fault of the target financial service, and the target financial service is trained based on a preset state space, action space, and reward function, the state space is constructed based on historical fault information, and the action space is constructed based on the operation information of multiple preset operations.

[0019] To achieve the above objectives, according to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the above-described fault prediction method for financial services.

[0020] To achieve the above objectives, according to another aspect of this application, an electronic device is provided, comprising a memory storing an executable program; and a processor for running the program, wherein the program executes the aforementioned fault prediction method for financial services during runtime.

[0021] To achieve the above objectives, according to another aspect of this application, a computer program product is provided, including computer instructions that, when executed by a processor, implement the steps of the above-described fault prediction method for financial services.

[0022] In this embodiment, by acquiring service data of the target financial service over N time periods, effective acquisition of relevant data on the target financial service across multiple modalities and spatiotemporal scales is achieved. By processing the data features of the N service data using a fault prediction model, fault prediction results are obtained. The model can learn the comprehensive state of service operation based on relevant data across multiple modalities and spatiotemporal scales, especially the service behavior patterns at different time points, thereby improving the accuracy of fault prediction results and thus improving the accuracy of fault prediction for financial services.

[0023] Therefore, the method provided in this application achieves the goal of fault prediction based on multimodal and multi-temporal scale data of the target financial service, realizes the technical effect of improving the accuracy of fault prediction, and solves the technical problem of low accuracy of fault prediction of financial services in related technologies. Attached Figure Description

[0024] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0025] Figure 1 This is a hardware structure block diagram of a computer terminal provided according to an embodiment of this application;

[0026] Figure 2 This is a flowchart of a fault prediction method for financial services provided according to embodiments of this application;

[0027] Figure 3 This is a schematic diagram of a fault prediction device for financial services provided according to an embodiment of this application;

[0028] Figure 4 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0032] Example 1

[0033] According to an embodiment of this application, a method embodiment for predicting failures in financial services is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0034] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a fault prediction method in financial services is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor (MCU) or a field-programmable gate array (FPGA), etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output (I / O) interface, a Universal Serial Bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0035] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0036] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the fault prediction method for financial services in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned fault prediction method for financial services. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0037] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0038] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0039] Under the aforementioned operating environment, this application provides the following: Figure 2 The method for predicting failures in financial services is shown. Figure 2 This is a flowchart of a fault prediction method for financial services provided according to an embodiment of this application.

[0040] Step S201: Obtain service data of the target financial service in the business platform within N time ranges. The service data includes at least: service logs and service performance data, where N is a positive integer greater than 1.

[0041] Optionally, electronic devices, application systems, servers, and other devices can be used as the execution subject of this application. In this embodiment, the target processing system is used as the execution subject to execute the above-mentioned fault prediction method for financial services.

[0042] Optionally, the business platform is the business platform of a financial institution used to provide financial services to users. The target financial service refers to the specific financial service in the business platform that requires fault prediction, such as online payment, account management, and transaction clearing.

[0043] Optionally, the N time ranges can have the same duration. For example, if the duration of a time range is 1 hour, the N time ranges can be the previous 6 hours, i.e., N=6. Or, if the duration of a time range is 10 minutes, the N time ranges can be the previous 6 ten-minute intervals, N=6. The N time ranges must be consecutive, and the total time range formed by the N time ranges can end at the current time.

[0044] Optionally, service data may include at least service logs and service performance data. Service logs record various events, warnings, and error messages during service runtime; service performance data reflects the service's running status, such as CPU (Central Processing Unit) utilization, memory usage, and network traffic.

[0045] In an optional embodiment, the service data may further include a system topology diagram of the business platform, which at least represents the calling relationships between P financial services in the business platform, where the P financial services include the target financial service and P is a positive integer greater than 1.

[0046] Step S202: Perform feature extraction processing on the service data within each time range to obtain the data features of each service data.

[0047] Optionally, for each time range of service data, feature extraction is performed on the service data to obtain the data features of the service data. For example, different feature extraction methods are used to extract features from different types of service data (service logs, service performance data), and then the data features are determined based on the extracted features.

[0048] Step S203: Process the data features of N service data through the fault prediction model to obtain fault prediction results, wherein the fault prediction results represent the probability of different fault types occurring in the target financial service.

[0049] Optionally, the fault prediction model can be a deep learning model.

[0050] Optionally, the target processing system can input the data features of N service data into the fault prediction model, so that the fault prediction model can process the N service data to obtain the fault prediction result.

[0051] Step S204: Determine target information based on the fault prediction results, wherein the target information characterizes whether the target financial service has failed and the type of failure if the target financial service fails.

[0052] Optionally, the target processing system can determine whether the target financial service has failed based on the probability value in the fault prediction result by means of threshold comparison, and if it is determined that the target financial service has failed, determine the fault type of the target financial service in the case of failure based on the magnitude of the probability value.

[0053] In this embodiment, by acquiring service data of the target financial service over N time periods, effective acquisition of relevant data on the target financial service across multiple modalities and spatiotemporal scales is achieved. By processing the data features of the N service data using a fault prediction model, fault prediction results are obtained. The model can learn the comprehensive state of service operation based on relevant data across multiple modalities and spatiotemporal scales, especially the service behavior patterns at different time points, thereby improving the accuracy of fault prediction results and thus improving the accuracy of fault prediction for financial services.

[0054] Therefore, the method provided in this application achieves the goal of fault prediction based on multimodal and multi-temporal scale data of the target financial service, realizes the technical effect of improving the accuracy of fault prediction, and solves the technical problem of low accuracy of fault prediction of financial services in related technologies.

[0055] Optionally, in the fault prediction method for financial services provided in this application embodiment, feature extraction processing is performed on the service data within each time range to obtain the data features of each service data, including: for the service data within each time range, feature extraction is performed on the service logs through a word embedding model to obtain a first vector; based on a preset sliding window, the time range is divided into M time windows, and the target features of the service performance data in each time window are calculated, wherein the target features include at least one of the following types of features: mean, variance, skewness, kurtosis, first difference, second difference, where M is a positive integer greater than 1; the target features of the service performance data in the M time windows are processed by a time series feature extractor to obtain a second vector; and the data features of the service data are determined based on the first vector and the second vector.

[0056] Optionally, service logs from the service data are input into a word embedding model to obtain a first vector. In an optional embodiment, key entity identification (such as server name, error code, operation type) can be performed on the logs, and importance weights can be assigned to them to strengthen the semantics of key entities during the embedding process.

[0057] Optionally, based on a preset sliding window, the time range is divided into M time windows. For example, if the original time range is 1 hour, it can be divided into multiple 10-minute time windows according to the preset sliding window; if the original time range is 10 minutes, it can be divided into multiple 1-minute time windows according to the preset sliding window. For service performance data (such as CPU utilization, response time, etc.), within each time window, a target feature is calculated. For example, assuming the service performance data includes CPU utilization and memory utilization, at least one of the following types of features is calculated for CPU utilization within the time window: mean, variance, skewness, kurtosis, first difference, and second difference, to obtain the target sub-feature corresponding to CPU utilization; and at least one of the following types of features is calculated for memory utilization within the time window: mean, variance, skewness, kurtosis, first difference, and second difference, to obtain the target sub-feature corresponding to memory utilization. Thus, the target feature is composed of the target sub-feature corresponding to CPU utilization and the target sub-feature corresponding to memory utilization.

[0058] Optionally, the mean, variance, skewness, and kurtosis mentioned above can be understood as statistical characteristics, while the first-order difference and second-order difference can be understood as trend characteristics.

[0059] Optionally, after obtaining the target features for each time window, a time-series feature extractor is used to extract features from the target features of the service performance data across M time windows to obtain a second vector. The time-series feature extractor can be a time-series encoder.

[0060] Optionally, after obtaining the first vector and the second vector, the data features of the service data are determined based on the first vector and the second vector. For example, the first vector and the second vector are concatenated to obtain the data features of the service data.

[0061] It should be noted that by extracting semantic features (first vector) from service logs and statistical and trend features (second vector) from service performance data, the richness of data features can be effectively improved, thereby enhancing the accuracy of fault prediction.

[0062] Optionally, in the fault prediction method for financial services provided in the embodiments of this application, the service data further includes a system topology diagram of the business platform. The system topology diagram represents at least the calling relationship between P financial services in the business platform. The P financial services include the target financial service, where P is a positive integer greater than 1. The process of determining the data features of the service data based on the first vector and the second vector includes: processing the system topology diagram through a graph neural network to obtain a third vector; and determining the data features of the service data based on the first vector, the second vector, and the third vector.

[0063] Optionally, the system topology graph should at least represent the calling relationships between P financial services in the business platform. For example, nodes in the system topology graph represent financial services, and edges represent the calling relationships between financial services.

[0064] In one optional embodiment, the system topology graph, in addition to representing the calling relationships between P financial services in the business platform, can also represent the deployment relationships of financial services on physical devices. For example, nodes in the system topology graph represent financial services or physical devices, and edges represent the calling relationships between financial services or the deployment relationships between financial services and physical devices.

[0065] Optionally, if the service data also includes a system topology diagram of the business platform, the target processing system can input the system topology diagram into a graph neural network, process the system topology diagram through the graph neural network, and obtain a third vector.

[0066] In an optional embodiment, after obtaining the third vector, the first vector, the second vector, and the third vector are concatenated to obtain the data features of the service data.

[0067] In an optional embodiment, after obtaining the third vector, the first, second, and third vectors are weighted and fused through an attention fusion layer to obtain the data features of the service data. For example, a small neural network is designed with inputs of modal vectors (i.e., the first, second, and third vectors). The neural network has pre-set attention weights for each modality, and the data features are obtained by weighted summation of the modal vectors based on the attention weights of each modality.

[0068] It should be noted that by introducing and processing the system topology diagram, not only are the characteristics of service logs and performance data considered, but also the topology information is taken into account. This enables the identification of dependencies between services and fault propagation patterns, thereby further improving the accuracy of fault prediction.

[0069] Optionally, in the fault prediction method for financial services provided in this application embodiment, the fault prediction model is obtained in the following ways: A target training sample set is obtained, wherein the training samples in the target training sample set are sample service data of the sample financial services, and the true labels of the training samples represent the fault types of the sample financial services; the target training sample set is processed by an initial fault prediction model to obtain the sample prediction results output by the output layer of the initial fault prediction model and the intermediate features output by the first processing layer; the fault description information of the training samples is processed by a fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model, wherein the fault analysis model is a large language model, and the fault analysis model is used to determine the cause and scope of fault impact based on the fault description information; a loss function value is determined based on the difference information between the sample prediction results and the true labels, and the difference information between the fault features and the intermediate features; and the model parameters of the initial fault prediction model are optimized based on the loss function value until the loss function value is less than a preset value, thus obtaining the fault prediction model.

[0070] Optionally, the training samples in the target training sample set are sample service data of the sample financial services, and the true labels of the training samples represent the fault types of the sample financial services. Specifically, a training sample is the sample service data of a sample financial service over N historical time ranges, that is, a training sample is a time-feature sequence S=V_fused_t1, V_fused_t2, ..., V_fused_tn).

[0071] Optionally, the target training sample set is processed using the initial fault prediction model. The model outputs a prediction result for the fault type of the sample financial service (i.e., the sample prediction result). During model processing, the first processing layer generates intermediate features. These features can be understood as a byproduct of the model's "thinking" process, containing its understanding and abstraction of the service data. The first processing layer is a specific processing layer preceding the output layer of the initial fault prediction model; for example, it could be a convolutional layer, pooling layer, etc. Therefore, when processing the target training sample set using the initial fault prediction model, in addition to obtaining the sample prediction result, the intermediate features output by the first processing layer are also acquired.

[0072] Optionally, the fault analysis model is a large language model, used to determine the cause and scope of the fault based on fault description information. The target processing system can fine-tune the large language model using a first training sample set to obtain the fault analysis model. For example, the training samples in the first training sample set are manually written fault reports, which include information such as fault phenomena, fault type, fault occurrence time, fault cause, and scope of fault impact. The true labels of the training samples in the first training sample set can be the manually simplified scope of fault impact and fault cause. Through the first training sample set, the large language model can learn knowledge related to faults in financial services.

[0073] Optionally, the fault description information of the training samples is pre-set manually. The target processing system can input the fault description information of the training samples into the fault analysis model, and process the fault description information of the training samples through the fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model. This vector integrates the semantic information of the fault description, including but not limited to the specific manifestation of the fault, possible root causes, and the scope of service impact. The second processing layer is a specific processing layer located before the output layer of the fault analysis model. For example, the second processing layer can be the last hidden layer or the penultimate layer.

[0074] Optionally, the target processing system can calculate the classification error (e.g., cross-entropy loss) between the sample prediction results of the initial fault prediction model and the true fault type labels of the training samples. This reflects the accuracy of the model's prediction. It can also compare the distance (e.g., mean squared error or cosine similarity loss) between the fault feature vectors generated by the fault analysis model and the intermediate feature vectors output by the initial model. This latter part of the loss indicates the degree of consistency between the two models' understanding of the data. Thus, the loss function value is determined based on the classification error and vector distance.

[0075] Optionally, if the loss function value is greater than a preset threshold, the model parameters of the initial fault prediction model are optimized based on the loss function value, and the step of processing the target training sample set with the updated initial fault prediction model to redetermine the loss function value is repeated until the loss function value is less than a preset value, and the latest obtained initial fault prediction model is determined as the fault prediction model.

[0076] In an optional embodiment, the fault prediction model may be a time-series prediction model.

[0077] It should be noted that, through the above methods, the fault prediction model, while learning to predict faults, can absorb the fault analysis model's deep understanding of fault context and semantic relationships. This allows it to identify more abstract and semantically dependent fault patterns (such as the logical chain of "service A timeout causing service B queue backlog"). In other words, through knowledge distillation (i.e., comparing fault features with intermediate features), the model can understand and capture the complex causal relationships and logical chains behind fault patterns. The fault prediction model can internally contain the knowledge of the fault analysis model, exhibiting better robustness and adaptability when facing unknown future faults, thereby improving fault prediction accuracy.

[0078] Optionally, in the fault prediction method for financial services provided in this application embodiment, determining the loss function value based on the difference information between the sample prediction result and the true label, and the difference information between the fault feature and the intermediate feature, includes: determining a first loss value based on a first sub-loss function, the sample prediction result, and the true label; determining a second loss value based on a second sub-loss function, the fault feature, and the intermediate feature; and determining the loss function value based on the first loss value, the second loss value, a first coefficient associated with the first sub-loss function, and a second coefficient associated with the second sub-loss function.

[0079] Optionally, the first sub-loss function can be a classification loss function, such as the cross-entropy loss function, used to measure the difference between the model's predicted results and the true fault type labels. The target processing system can determine the first loss value based on the first sub-loss function, the sample prediction results, and the true labels.

[0080] Optionally, the second sub-loss function can be a knowledge distillation loss function, such as mean squared error or cosine similarity loss, used to measure the similarity between the intermediate features of the model and the fault features output by the fault analysis model. The target processing system can determine the second loss value based on the second sub-loss function, the fault features, and the intermediate features.

[0081] Optionally, the target processing system can calculate the product between the first coefficient and the first loss value to obtain a first value, calculate the product between the second coefficient and the second loss value to obtain a second value, and then add the first value and the second value to obtain the loss function value.

[0082] Optionally, the sum of the first coefficient and the second coefficient is 1, and the first coefficient and the second coefficient can be preset values.

[0083] It should be noted that by designing a composite loss function, the improvement in prediction accuracy is combined with the deepening of knowledge distillation, achieving a two-pronged optimization of the fault prediction model. This approach not only improves the accuracy of the model in predicting fault types but also enhances the model's understanding of the underlying logical links of faults. This is particularly valuable for predicting complex faults caused by multiple factors, thereby improving the accuracy of fault prediction.

[0084] Optionally, in the fault prediction method for financial services provided in this application embodiment, determining target information based on the fault prediction result includes: determining whether there is a target probability value in the fault prediction result, wherein the target probability value is a probability value greater than a preset probability; if there is a target probability value in the fault prediction result, determining that the target information represents a fault in the target financial service, and determining the fault type corresponding to the highest target probability value as the fault type of the target financial service; if there is no target probability value in the fault prediction result, determining that the target information represents no fault in the target financial service.

[0085] Optionally, if the probability of a certain type of fault predicted by the model exceeds a preset threshold, then this type of fault is considered likely to occur. Therefore, the target processing system can first determine whether the fault prediction result has a target probability value.

[0086] Optionally, if a fault type exceeding a preset probability threshold is found in the prediction results, then it can be determined that the target financial service has a fault risk. In this case, the fault type corresponding to the highest target probability value is determined as the fault type of the target financial service.

[0087] Optionally, if after checking the predicted probabilities of all fault types, it is found that none of the probability values ​​exceed the preset threshold, then it can be determined that the target financial service is currently operating stably and has not experienced any faults.

[0088] It should be noted that determining target information by using threshold comparison improves the accuracy of the determined target information.

[0089] Optionally, in the fault prediction method for financial services provided in this application embodiment, when a target probability value exists in the target financial service, after determining the target information based on the fault prediction result, the method further includes: obtaining system resource information of the target financial service, and composing fault information based on the highest target probability value, the fault type of the target financial service, and the system resource information of the target financial service; processing the fault information through a fault handling model to obtain target operation information, wherein the target operation information includes operation information for repairing the fault of the target financial service, the target financial service is trained based on a preset state space, action space, and reward function, the state space is constructed based on historical fault information, and the action space is constructed based on the operation information of multiple preset operations.

[0090] Optionally, system resource information refers to the current operating status and resource usage of the target financial service, including but not limited to key indicators such as CPU utilization, memory usage, disk I / O, network bandwidth, and service response time.

[0091] Optionally, the fault handling model can be a reinforcement learning-based model. The target processing system can input system resource information into the fault handling model to obtain target operation information, which includes operation information for repairing faults in the target financial service. After obtaining the target operation information, fault repair can be performed on the target financial service based on the target operation information (or fault prevention if the fault has not yet actually occurred).

[0092] Optionally, the target financial service is trained based on a preset state space, action space, and reward function. The state space is constructed based on historical fault information, which may include the highest probability value in the fault prediction results of the fault prediction model for the sample financial service, the fault type corresponding to the highest probability value, and the system resource information of the sample financial service. The action space is constructed based on the operation information of multiple preset operations, which may include restart, expansion, switching, alarm suppression, etc., and the operation information may refer to the operation identifier code of the operation.

[0093] Optionally, the reward function is the core driver of the agent's learning. Its design needs to balance multiple objectives; for example, the composition of the reward function is shown below:

[0094] R1: Successfully prevent or quickly repair predicted faults (based on matching prediction model alarms with actual fault occurrences).

[0095] R2: A fault has occurred.

[0096] R3: Unnecessary maintenance actions were performed.

[0097] R4: The cost of the operation and maintenance actions themselves (such as resource consumption and service interruption time).

[0098] R5: Alarm Storm (Suppresses valid alarms or generates excessive redundant alarms).

[0099] R6 (optional): Improves overall system stability and performance.

[0100] For example, an alternative reward function R can be expressed as R = w1R1 + w2R2 + w3R3 + w4R4 + w5R5 + w6R6. Here, w1, w2, w3, w4, w5, and w6 represent weights.

[0101] Optionally, during the training process, the fault handling model involves the agent exploring a simulated environment (based on historical data playback or a system model) or a carefully controlled production environment. The agent selects action A_t in state S_t, the environment executes the action and transitions to a new state S_t+1, awarding a reward R_t. The agent updates the policy network π(A|S) and the value function network V(S) by maximizing the cumulative expected discounted reward. The trained policy network π is used for online decision-making. When the actual state (i.e., the fault information mentioned above) is received, the policy network outputs the optimal action A (i.e., the target operation information) to be executed.

[0102] In an optional embodiment, the target processing system can utilize feedback from actual operation and maintenance results to continuously improve the accuracy of the prediction model and the effectiveness of the strategy. For example, it records the prediction records of the fault prediction model, the action records determined by the fault processing model, and the actual result records determined by manual assistance during actual application (e.g., whether a fault actually occurred, the time of the fault occurrence, the scope of impact, the actual repair actions, the repair time, and changes in system state). The prediction records and the corresponding result records (labeled with real fault tags) are used as new samples, and the fault prediction model is incrementally trained periodically (e.g., daily / weekly) or when a decline in the performance of the fault prediction model is detected.

[0103] Optionally, the action records from the actual application process, the state S_t before execution, the state S_t+1 after execution, and the actual reward R_t obtained can be used as new experience samples (S_t, A_t, R_t, S_t+1). Periodically or when the policy performance is poor, the newly collected experience samples are used to update the RL policy network π and the value function network V.

[0104] Optionally, confirmed failure cases, along with their complete data (logs, metrics, operations, results) and final analysis reports, can be added to the case library. The failure analysis model should be incrementally fine-tuned periodically using the updated case library to keep the knowledge up-to-date.

[0105] In an optional embodiment, the target processing system can be configured with a visualization dashboard to display real-time system health status, fault prediction results, key resource metrics, current alarm / action queues, etc. Operations personnel can input natural language (e.g., "Why was the database response slow last night?"), and the fault analysis model will retrieve similar historical cases, related logs / metrics, generate explanatory answers, and link to the original data. The target processing system can also display the action A recommended by the policy engine in the current state, along with its estimated reward / effect description. Operations personnel can view details, confirm execution, or manually select other actions.

[0106] It should be noted that the above method enables the automatic determination of fault repair operation suggestions, reducing the delay and uncertainty of manual decision-making, thereby accelerating the fault recovery process.

[0107] Therefore, the method provided in this application achieves the goal of fault prediction based on multimodal and multi-temporal scale data of the target financial service, realizes the technical effect of improving the accuracy of fault prediction, and solves the technical problem of low accuracy of fault prediction of financial services in related technologies.

[0108] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0109] Example 2

[0110] This application also provides a fault prediction device for financial services. It should be noted that the fault prediction device for financial services in this application can be used to execute the fault prediction method for financial services provided in this application. The following describes the fault prediction device for financial services provided in this application.

[0111] According to embodiments of this application, an apparatus for implementing the fault prediction method for the aforementioned financial services is also provided, such as... Figure 3 As shown, the device includes:

[0112] The first acquisition module 301 is used to acquire service data of the target financial service in the business platform within N time ranges, wherein the service data includes at least: service logs and service performance data, and N is a positive integer greater than 1;

[0113] The first processing module 302 is used to perform feature extraction processing on the service data in each time range to obtain the data features of each service data.

[0114] The second processing module 303 is used to process the data features of N service data through a fault prediction model to obtain fault prediction results, wherein the fault prediction results represent the probability of different fault types occurring in the target financial service.

[0115] The first determining module 304 is used to determine target information based on the fault prediction results, wherein the target information characterizes whether the target financial service has failed and the type of failure in the event of a failure.

[0116] In this embodiment, by acquiring service data of the target financial service over N time periods, effective acquisition of relevant data on the target financial service across multiple modalities and spatiotemporal scales is achieved. By processing the data features of the N service data using a fault prediction model, fault prediction results are obtained. The model can learn the comprehensive state of service operation based on relevant data across multiple modalities and spatiotemporal scales, especially the service behavior patterns at different time points, thereby improving the accuracy of fault prediction results and thus improving the accuracy of fault prediction for financial services.

[0117] Therefore, the method provided in this application achieves the goal of fault prediction based on multimodal and multi-temporal scale data of the target financial service, realizes the technical effect of improving the accuracy of fault prediction, and solves the technical problem of low accuracy of fault prediction of financial services in related technologies.

[0118] Optionally, in the fault prediction device for financial services provided in this application embodiment, the first processing module further includes: a first extraction submodule, used to extract features from service logs using a word embedding model for service data within each time range to obtain a first vector; a calculation submodule, used to divide the time range into M time windows based on a preset sliding window, and calculate the target features of service performance data in each time window, wherein the target features include at least one of the following types of features: mean, variance, skewness, kurtosis, first difference, second difference, where M is a positive integer greater than 1; a second extraction submodule, used to process the target features of service performance data in the M time windows using a time-series feature extractor to obtain a second vector; and a first determination submodule, used to determine the data features of service data based on the first vector and the second vector.

[0119] Optionally, in the fault prediction device for financial services provided in the embodiments of this application, the first determining submodule further includes: a processing unit, used to process the system topology graph through a graph neural network to obtain a third vector; and a determining unit, used to determine the data features of the service data based on the first vector, the second vector, and the third vector.

[0120] Optionally, in the financial service fault prediction device provided in this application embodiment, the financial service fault prediction device further includes: a second acquisition module, used to acquire a target training sample set, wherein the training samples in the target training sample set are sample service data of the sample financial service, and the real labels of the training samples represent the fault type of the sample financial service; a third processing module, used to process the target training sample set through an initial fault prediction model to obtain the sample prediction result output by the output layer of the initial fault prediction model and the intermediate features output by the first processing layer; a fourth processing module, used to process the fault description information of the training samples through a fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model, wherein the fault analysis model is a large language model, and the fault analysis model is used to determine the fault cause and the fault impact range based on the fault description information; and a second determination module, used to determine the loss function value based on the difference information between the sample prediction result and the real label, and the difference information between the fault features and the intermediate features, and optimize the model parameters of the initial fault prediction model based on the loss function value until the loss function value is less than a preset value, thereby obtaining the fault prediction model.

[0121] Optionally, in the fault prediction device for financial services provided in the embodiments of this application, the second determining module further includes: a second determining submodule, used to determine a first loss value based on a first sub-loss function, sample prediction results, and true labels; a third determining submodule, used to determine a second loss value based on a second sub-loss function, fault features, and intermediate features; and a fourth determining submodule, used to determine a loss function value based on the first loss value, the second loss value, a first coefficient associated with the first sub-loss function, and a second coefficient associated with the second sub-loss function.

[0122] Optionally, in the fault prediction device for financial services provided in the embodiments of this application, the first determining module further includes: a judging submodule, used to judge whether there is a target probability value in the fault prediction result, wherein the target probability value is a probability value greater than a preset probability; a fifth determining submodule, used to determine that the target information represents a fault in the target financial service when there is a target probability value in the fault prediction result, and to determine the fault type corresponding to the highest target probability value as the fault type of the target financial service; and a sixth determining submodule, used to determine that the target information represents no fault in the target financial service when there is no target probability value in the fault prediction result.

[0123] Optionally, in the fault prediction device for financial services provided in this application embodiment, the fault prediction device for financial services further includes: a third acquisition module, used to acquire system resource information of the target financial service, and to compose fault information based on the highest target probability value, the fault type of the target financial service, and the system resource information of the target financial service; and a fifth processing module, used to process the fault information through a fault processing model to obtain target operation information, wherein the target operation information includes operation information for repairing the fault of the target financial service, the target financial service is trained based on a preset state space, action space, and reward function, the state space is constructed based on historical fault information, and the action space is constructed based on the operation information of multiple preset operations.

[0124] It should be noted that the first acquisition module 301, the first processing module 302, the second processing module 303, and the first determination module 304 mentioned above correspond to steps S201 to S204 in Embodiment 1. The four modules and the corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0125] Example 3

[0126] Embodiments of this application may provide an electronic device. Figure 4 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (Only one is shown) processor 1002, memory 1004, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0127] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0128] The processor can access information and applications stored in memory via a transmission device to perform the following steps: acquiring service data of the target financial service within N time ranges from the business platform, wherein the service data includes at least: service logs and service performance data, and N is a positive integer greater than 1; performing feature extraction processing on the service data within each time range to obtain data features for each service data; processing the data features of the N service data using a fault prediction model to obtain fault prediction results, wherein the fault prediction results characterize the probability of different fault types occurring in the target financial service; and determining target information based on the fault prediction results, wherein the target information characterizes whether the target financial service has experienced a fault, and the type of fault in the event of a fault.

[0129] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: For service data within each time range, feature extraction is performed on the service logs using a word embedding model to obtain a first vector; based on a preset sliding window, the time range is divided into M time windows, and the target features of the service performance data in each time window are calculated, wherein the target features include at least one of the following types of features: mean, variance, skewness, kurtosis, first difference, second difference, where M is a positive integer greater than 1; the target features of the service performance data in the M time windows are processed by a time-series feature extractor to obtain a second vector; the data features of the service data are determined based on the first vector and the second vector.

[0130] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: The fault prediction method for financial services also includes: processing the system topology graph through a graph neural network to obtain a third vector; determining the data features of the service data based on the first vector, the second vector and the third vector.

[0131] The processor can also invoke information and applications stored in the memory via a transmission device to perform the following steps: acquiring a target training sample set, wherein the training samples in the target training sample set are sample service data of sample financial services, and the true labels of the training samples represent the fault types of sample financial services; processing the target training sample set through an initial fault prediction model to obtain the sample prediction results output by the output layer of the initial fault prediction model and the intermediate features output by the first processing layer; processing the fault description information of the training samples through a fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model, wherein the fault analysis model is a large language model, and the fault analysis model is used to determine the cause of the fault and the scope of the fault impact based on the fault description information; determining the loss function value based on the difference information between the sample prediction results and the true labels, and the difference information between the fault features and the intermediate features, and optimizing the model parameters of the initial fault prediction model based on the loss function value until the loss function value is less than a preset value, thereby obtaining the fault prediction model.

[0132] The processor can also call the information and application program stored in the memory through the transmission device to perform the following steps: determine a first loss value based on the first sub-loss function, the sample prediction result and the true label; determine a second loss value based on the second sub-loss function, the fault features and the intermediate features; and determine the loss function value based on the first loss value, the second loss value, the first coefficient associated with the first sub-loss function and the second coefficient associated with the second sub-loss function.

[0133] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: determine whether there is a target probability value in the fault prediction result, wherein the target probability value is a probability value that is greater than a preset probability; if there is a target probability value in the fault prediction result, determine that the target information represents a fault in the target financial service, and determine the fault type corresponding to the highest target probability value as the fault type of the target financial service; if there is no target probability value in the fault prediction result, determine that the target information represents no fault in the target financial service.

[0134] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: obtain the system resource information of the target financial service, and compose fault information based on the highest target probability value, the fault type of the target financial service, and the system resource information of the target financial service; process the fault information through the fault handling model to obtain target operation information, wherein the target operation information includes operation information for repairing the fault of the target financial service, the target financial service is trained based on a preset state space, action space and reward function, the state space is constructed based on historical fault information, and the action space is constructed based on the operation information of multiple preset operations.

[0135] In this embodiment, by acquiring service data of the target financial service over N time periods, effective acquisition of relevant data on the target financial service across multiple modalities and spatiotemporal scales is achieved. By processing the data features of the N service data using a fault prediction model, fault prediction results are obtained. The model can learn the comprehensive state of service operation based on relevant data across multiple modalities and spatiotemporal scales, especially the service behavior patterns at different time points, thereby improving the accuracy of fault prediction results and thus improving the accuracy of fault prediction for financial services.

[0136] Therefore, the method provided in this application achieves the goal of fault prediction based on multimodal and multi-temporal scale data of the target financial service, realizes the technical effect of improving the accuracy of fault prediction, and solves the technical problem of low accuracy of fault prediction of financial services in related technologies.

[0137] Those skilled in the art will understand that Figure 4 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 4 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 4 The different configurations shown.

[0138] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0139] Example 4

[0140] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the fault prediction method for financial services provided in Embodiment 1.

[0141] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0142] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing steps of a fault prediction method for financial services.

[0143] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0144] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0145] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0146] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0148] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0149] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for predicting failures in financial services, characterized in that, include: Obtain service data of the target financial service in the business platform within N time ranges, wherein the service data includes at least: service logs and service performance data, and N is a positive integer greater than 1; Feature extraction is performed on the service data for each time range to obtain the data features of each service data; The fault prediction model processes the data features of N service data to obtain fault prediction results, wherein the fault prediction results characterize the probability of different fault types occurring in the target financial service. Target information is determined based on the fault prediction results, wherein the target information characterizes whether the target financial service has failed, and the type of failure of the target financial service if it has failed.

2. The method according to claim 1, characterized in that, Feature extraction is performed on the service data for each time range to obtain the data features of each service data, including: For the service data within each time range, feature extraction is performed on the service logs using a word embedding model to obtain a first vector; Based on a preset sliding window, the time range is divided into M time windows, and the target features of the service performance data in each time window are calculated. The target features include at least one of the following types of features: mean, variance, skewness, kurtosis, first difference, second difference, where M is a positive integer greater than 1. The service performance data is processed using a temporal feature extractor over M time windows to obtain a second vector; The data characteristics of the service data are determined based on the first vector and the second vector.

3. The method according to claim 2, characterized in that, The service data also includes a system topology diagram of the business platform, which represents at least the calling relationships between P financial services in the business platform. The P financial services include the target financial service, where P is a positive integer greater than 1. The data characteristics of the service data determined based on the first vector and the second vector include: The system topology graph is processed by a graph neural network to obtain a third vector; The data characteristics of the service data are determined based on the first vector, the second vector, and the third vector.

4. The method according to claim 1, characterized in that, The fault prediction model is obtained in the following way: Obtain a target training sample set, wherein the training samples in the target training sample set are sample service data of sample financial services, and the true labels of the training samples represent the fault types of the sample financial services; The target training sample set is processed by the initial fault prediction model to obtain the sample prediction results output by the output layer of the initial fault prediction model and the intermediate features output by the first processing layer. The fault description information of the training samples is processed by the fault analysis model to obtain the fault features output by the second processing layer of the fault analysis model. The fault analysis model is a large language model and is used to determine the cause of the fault and the scope of its impact based on the fault description information. The loss function value is determined based on the difference between the sample prediction result and the true label, and the difference between the fault feature and the intermediate feature. The model parameters of the initial fault prediction model are then optimized based on the loss function value until the loss function value is less than a preset value, thus obtaining the fault prediction model.

5. The method according to claim 4, characterized in that, The loss function value is determined based on the difference between the sample prediction result and the true label, and the difference between the fault feature and the intermediate feature, including: The first loss value is determined based on the first sub-loss function, the sample prediction result, and the true label; The second loss value is determined based on the second sub-loss function, the fault characteristics, and the intermediate characteristics; The loss function value is determined based on the first loss value, the second loss value, the first coefficient associated with the first sub-loss function, and the second coefficient associated with the second sub-loss function.

6. The method according to claim 1, characterized in that, Based on the fault prediction results, target information is determined, including: Determine whether the fault prediction result has a target probability value, wherein the target probability value is a probability value that is greater than a preset probability. If a target probability value exists in the fault prediction result, the target information is determined to represent a fault in the target financial service, and the fault type corresponding to the highest target probability value is determined as the fault type of the target financial service. If the target probability value is not present in the fault prediction result, the target information is determined to indicate that the target financial service has not experienced a fault.

7. The method according to claim 6, characterized in that, If a target probability value exists in the target financial service, after determining the target information based on the fault prediction result, the method further includes: Obtain system resource information of the target financial service, and compose fault information based on the highest target probability value, the fault type of the target financial service, and the system resource information of the target financial service; The fault information is processed by a fault handling model to obtain target operation information, wherein the target operation information includes operation information for repairing the fault of the target financial service. The target financial service is trained based on a preset state space, action space and reward function. The state space is constructed based on historical fault information and the action space is constructed based on the operation information of multiple preset operations.

8. A fault prediction device for financial services, characterized in that, include: The first acquisition module is used to acquire service data of the target financial service in the business platform within N time ranges, wherein the service data includes at least: service logs and service performance data, and N is a positive integer greater than 1; The first processing module is used to perform feature extraction processing on the service data in each time range to obtain the data features of each service data. The second processing module is used to process the data features of N service data through a fault prediction model to obtain fault prediction results, wherein the fault prediction results represent the probability of different fault types occurring in the target financial service. The first determining module is used to determine target information based on the fault prediction result, wherein the target information characterizes whether the target financial service has failed and the type of failure of the target financial service in the event of failure.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium is located to perform the fault prediction method for financial services as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the fault prediction method for financial services according to any one of claims 1 to 7.

11. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the fault prediction method for financial services as described in any one of claims 1 to 7.

Images