A bank equipment failure early warning method and system based on machine learning

Abstract

The present invention discloses a bank equipment failure early warning method and system based on machine learning, which includes obtaining first training model information, obtaining first sample data from first bank equipment information, using SVM and naive Bayes algorithm to train the first training model; obtaining first original data information; inputting the first original data information into the first training model after training to obtain output information; obtaining first feature information; obtaining second classification result information; according to the second classification result information, if a first warning instruction needs to be issued, the first warning instruction is sent to a manual verification unit through an alarm interface to obtain a first verification result of manual investigation; and adjusting the first training model according to the first verification result and the first predicted classification result. The method solves the technical problem that the bank cannot or cannot accurately carry out active early warning for possible abnormalities in the prior art, thereby affecting the operation and maintenance efficiency.

Description

Technical Field

[0001] The present invention relates to the field related to bank equipment failure early warning, and in particular to a bank equipment failure early warning method and system based on machine learning. Background Art

[0002] In banking systems, equipment abnormality alarms play a vital role in abnormality control. Currently, there are two main ways to generate alarms for banking equipment and software systems: (1) monitoring system abnormality logs. If the system captures an abnormality, it will trigger a text message or email alarm; (2) for abnormalities that cannot be captured or hardware equipment failures, manual discovery is required.

[0003] However, the inventors of this application found that the above technology has at least the following technical problems in the process of implementing the technical solution of the invention in the embodiment of this application:

[0004] There is a technical problem in the existing technology that the bank cannot or cannot accurately provide active warnings for possible abnormalities, thereby affecting the operation and maintenance efficiency. Summary of the invention

[0005] The embodiments of the present application provide a bank equipment fault warning method and system based on machine learning, which solves the technical problem in the prior art that the bank cannot or cannot accurately provide active warnings for possible abnormalities, thereby affecting the operation and maintenance efficiency. The present application achieves the technical effect of providing active and accurate warnings for possible abnormalities, reducing the impact of faults on the system, and improving the operation and maintenance efficiency.

[0006] In view of the above problems, an embodiment of the present application provides a banking equipment failure warning method and system based on machine learning.

[0007] In a first aspect, an embodiment of the present application provides a bank equipment failure warning method based on machine learning, the method is applied to a risk assessment system, and the system includes a model training unit, a raw data monitoring unit, a feature engineering unit, a classification algorithm unit, a warning unit, and a manual verification unit, wherein the method includes: obtaining first training model information, wherein the first training model is a binary classification prediction model; obtaining first sample data from first bank equipment information, wherein the first sample data includes negative sample data and positive sample data; according to the first sample data, using SVM and naive Bayes algorithm to train the first training model; obtaining first raw data information through the raw data monitoring unit; converting the first raw data information into The information is input into the first training model after training is completed to obtain the output information of the first training model, wherein the output information is the first prediction classification result information; according to the first original data information, the first feature information is obtained through the feature engineering unit; according to the first feature information, the second classification result information is obtained; according to the second classification result information, it is determined whether to issue a first warning instruction; if it is necessary to issue a first warning instruction, the first warning instruction is sent to the manual verification unit through the alarm interface, so that the administrator performs manual investigation after receiving the first warning instruction; the first verification result of the manual investigation is obtained; according to the first verification result and the first prediction classification result, the first training model is adjusted.

[0008] On the other hand, the present application also provides a bank equipment failure warning system based on machine learning, the system comprising: a first acquisition unit, the first acquisition unit is used to obtain first training model information, wherein the first training model is a binary classification prediction model; a second acquisition unit, the second acquisition unit is used to obtain first sample data from the first bank equipment information, wherein the first sample data includes negative sample data and positive sample data; a first training unit, the first training unit is used to train the first training model according to the first sample data using SVM and naive Bayes algorithm; a third acquisition unit, the third acquisition unit is used to obtain first raw data information through a raw data monitoring unit; a first input unit, the first input unit is used to input the first raw data information into the first training model after training, and obtain output information of the first training model, wherein the input The output information is the first prediction classification result information; the fourth obtaining unit, the fourth obtaining unit is used to obtain the first feature information through the feature engineering unit according to the first original data information; the fifth obtaining unit, the fifth obtaining unit is used to obtain the second classification result information according to the first feature information; the first determining unit, the first determining unit is used to determine whether to issue a first warning instruction according to the second classification result information; the sixth obtaining unit, the sixth obtaining unit is used to send the first warning instruction to the manual verification unit through the alarm interface if the first warning instruction needs to be issued, so that the administrator can perform manual investigation after receiving the first warning instruction; the seventh obtaining unit, the seventh obtaining unit is used to obtain the first verification result of the manual investigation; the first adjustment unit, the first adjustment unit is used to adjust the first training model according to the first verification result and the first prediction classification result.

[0009] In a third aspect, the present invention provides a bank equipment failure warning system based on machine learning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the program.

[0010] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0011] Because the first training model is trained according to the first sample data using SVM and Naive Bayes algorithm, the first original data is input into the first training model to obtain output information of the first training model, the first original data is used through a feature engineering unit to obtain second classification result information, and it is determined whether to perform manual investigation according to the second classification result information, and the first training model is adjusted according to the manual investigation result, so as to obtain a more accurate first training model to process the input data, thereby achieving active and accurate early warning of possible anomalies, reducing the impact of failures on the system, and achieving the technical effect of improving operation and maintenance efficiency.

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of a bank equipment failure early warning method based on machine learning according to an embodiment of the present application;

[0014] Figure 2 This is a structural diagram of a bank equipment failure early warning system based on machine learning according to an embodiment of the present application;

[0015] Figure 3 It is a schematic diagram of the structure of an exemplary electronic device according to an embodiment of the present application.

[0016] Explanation of the figure marks: first acquisition unit 11, second acquisition unit 12, first training unit 13, third acquisition unit 14, first input unit 15, fourth acquisition unit 16, fifth acquisition unit 17, first determination unit 18, sixth acquisition unit 19, seventh acquisition unit 20, first adjustment unit 21, bus 300, receiver 301, processor 302, transmitter 303, memory 304, bus interface 306. DETAILED DESCRIPTION

[0017] The embodiment of the present application solves the technical problem in the prior art that the bank cannot or cannot accurately provide active warnings for possible abnormalities that may occur, thereby affecting the operation and maintenance efficiency, by providing a bank equipment fault warning method and system based on machine learning, so as to achieve the technical effect of providing active and accurate warnings for possible abnormalities, reducing the impact of faults on the system, and improving the operation and maintenance efficiency. Below, the example embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described here.

[0018] Application Overview

[0019] In the banking system, the abnormal alarm of the equipment plays a vital role in abnormal control. At present, there are two main ways to warn the bank's equipment and software systems: (1) monitoring the system abnormality log, and triggering SMS or email alarms if the system captures an abnormality; (2) for abnormalities that cannot be captured or hardware equipment failures, manual active discovery. However, there is a technical problem in the existing technology that the bank's possible abnormalities cannot be proactively warned or cannot be accurately warned, thereby affecting the efficiency of operation and maintenance.

[0020] In response to the above technical problems, the overall idea of ​​the technical solution provided by this application is as follows:

[0021] The embodiment of the present application provides a bank equipment failure early warning method based on machine learning, the method is applied to a risk assessment system, and the system includes a model training unit, a raw data monitoring unit, a feature engineering unit, a classification algorithm unit, an early warning unit, and a manual verification unit, wherein the method includes: obtaining first training model information, wherein the first training model is a binary classification prediction model; obtaining first sample data from first bank equipment information, wherein the first sample data includes negative sample data and positive sample data; according to the first sample data, using SVM and naive Bayes algorithm to train the first training model; obtaining first raw data information through the raw data monitoring unit; inputting the first raw data information into the system; The first training model is input into the first training model after training, and the output information of the first training model is obtained, wherein the output information is the first prediction classification result information; according to the first original data information, the first feature information is obtained through the feature engineering unit; according to the first feature information, the second classification result information is obtained; according to the second classification result information, it is determined whether to issue a first warning instruction; if it is necessary to issue a first warning instruction, the first warning instruction is sent to the manual verification unit through the alarm interface, so that the administrator can perform manual investigation after receiving the first warning instruction; the first verification result of the manual investigation is obtained; according to the first verification result and the first prediction classification result, the first training model is adjusted.

[0022] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0023] Embodiment 1

[0024] like Figure 1As shown, the embodiment of the present application provides a bank equipment failure early warning method based on machine learning, the method is applied to a risk assessment system, and the system includes a model training unit, a raw data monitoring unit, a feature engineering unit, a classification algorithm unit, an early warning unit, and a manual verification unit, wherein the method includes:

[0025] Step S100: obtaining first training model information, wherein the first training model is a binary classification prediction model;

[0026] Specifically, the first training model is a model that can be continuously supervised and trained for correction, and the first training model is a binary prediction model. The binary prediction model is a prediction model based on machine learning of a confusion matrix, which can obtain more accurate prediction results. The first training model is obtained by using SVM combined with a naive Bayes algorithm to train a binary prediction model.

[0027] Step S200: obtaining first sample data from first bank device information, wherein the first sample data includes negative sample data and positive sample data;

[0028] Specifically, the negative sample data are error logs of major systems of the bank, for example, logs within 10 hours before the abnormality occurs, and the positive sample data are logs during the normal operation of the system or equipment.

[0029] Step S300: training the first training model using SVM and naive Bayes algorithm according to the first sample data;

[0030] Specifically, through the above-mentioned first sample data, SVM combined with the naive Bayes algorithm is used to train the binary classification prediction model. According to the characteristics of the above two algorithms: SVM is used for prediction when there are a small number of samples; naive Bayes is used for prediction when the sample size is large. By combining SVM and the naive Bayes algorithm, the technical effect of improving the accuracy of model classification is achieved, which lays a solid foundation for the subsequent accurate prediction of equipment failures.

[0031] Step S400: obtaining first original data information through the original data monitoring unit;

[0032] Specifically, the original data monitoring unit is a unit that performs log monitoring and obtains original data, and obtains first original data information from the monitored log.

[0033] Step S500: inputting the first original data information into the first training model after training, and obtaining output information of the first training model, wherein the output information is first prediction classification result information;

[0034] Specifically, the first original data information is input into the first training model obtained after sample data training to obtain the output result of the first training model, and the output result includes the classification result after the first training model predicts the first original data information, that is, the first predicted classification result information.

[0035] Step S600: obtaining first feature information through the feature engineering unit according to the first original data information;

[0036] Step S700: obtaining second classification result information according to the first feature information;

[0037] Specifically, the feature engineering unit is a unit for feature comparison and screening. The screening process is to first perform data cleaning, vectorization, missing value supplementation, and deletion of features with small influencing factors, and use three methods: complete search, heuristic search, and random algorithm to screen out high-impact features to prepare for the next step of classification. The second classification result is obtained based on the first feature information, including inputting the feature information into the classification algorithm unit, using two classification algorithms, analyzing the features, and obtaining the second classification result information.

[0038] Step S800: determining whether to issue a first warning instruction according to the second classification result information;

[0039] Step S900: If a first warning instruction needs to be issued, the first warning instruction is sent to the manual checking unit through the alarm interface, so that the administrator performs manual checking after receiving the first warning instruction;

[0040] Specifically, according to the output result after the feature ratio analysis, that is, the second classification result information, it is determined whether the second classification result needs to issue a first warning instruction. When the second classification result is that the first original data information may be abnormal, the first warning instruction is obtained. According to the first warning instruction, it is sent to the manual verification unit through the alarm interface. The warning instruction can be issued through the early warning unit. This early warning unit is mainly used to send an alarm function after the system predicts an abnormality. Generally, banks have their own alarm interfaces such as SMS and email, which are directly called after configuring the corresponding alarm content. The manual verification unit is a unit that manually checks the feature information of the first original data according to the warning instruction to verify the accuracy of the calculation result.

[0041] Step S1000: Obtaining the first verification result of the manual investigation;

[0042] Step S1100: Adjust the first training model according to the first verification result and the first prediction classification result.

[0043] Specifically, the first training model is adjusted according to the first verification result and the first prediction classification result, including: after receiving the abnormality predicted by the alarm, timely manual investigation is carried out, the accuracy and recall rate of the verification result are calculated, and finally the F1-Score is calculated to represent the comprehensive accuracy, and the trained model is retrained (fine-tuned) according to the F1-Score result: (1) If the accuracy meets the standard, it means that the weights of the training model are accurate, and only the training data needs to be increased; (2) If the accuracy does not meet the standard, it is necessary to fine-tune the model training parameters according to the prediction deviation, and retrain the first training model, so as to finally obtain a more accurate first training model to predict the original data, actively and accurately warn of possible abnormalities, reduce the impact of failures on the system, and achieve the technical effect of improving operation and maintenance efficiency.

[0044] Furthermore, the step S200 of obtaining the first sample data from the first bank device information in the embodiment of the present application further includes:

[0045] Step S210: Obtain a first preset time period;

[0046] Step S220: obtaining first abnormal event information according to the first preset time period;

[0047] Step S230: obtaining a second preset time period according to the first abnormal event information, wherein the second preset time period is before the first abnormal event occurs;

[0048] Step S240: obtaining the negative sample data according to the second preset time period;

[0049] Step S250: obtaining a third preset time period;

[0050] Step S260: obtaining the positive sample data according to the third preset time period;

[0051] Step S270: Obtain the first sample data according to the negative sample data and the positive sample data.

[0052] Specifically, in order to ensure the training accuracy of the binary classification model, the training sample data needs to be based on the confusion matrix. At this time, there is a certain requirement for the training sample data, that is, it contains positive sample data and negative sample data. The data is segmented by time, the first abnormal event information within the first preset time period is obtained, the second preset time period before the first abnormal event information occurs is obtained, the data of this time period is obtained as negative sample data, and the third preset time period is obtained. The third preset time period is a normal time period, that is, the third time period is not adjacent to the above-mentioned time period with abnormal events. The data of the third time period is used as positive sample data, and the positive sample data and the negative sample data are integrated to obtain the first sample data.

[0053] Furthermore, the first feature information is obtained by the feature engineering unit according to the first original data information. Step S600 of the embodiment of the present application further includes:

[0054] Step S610: obtaining a first cleaning instruction;

[0055] Step S620: After cleaning the first original data according to the first cleaning instruction, first valid data is obtained;

[0056] Step S630: obtaining a first processing instruction;

[0057] Step S640: performing vectorized processing on the first valid data according to the first processing instruction.

[0058] Specifically, the first data cleaning instruction is obtained and the first raw data is cleaned, which is actually to judge the integrity and validity of the collected data. When the collected data is incomplete, that is, missing data, the missing data is deleted. If invalid data is detected, the invalid data is also deleted according to the first data cleaning instruction. By cleaning the first data, the technical effect of ensuring the integrity and reliability of the data is achieved. The first valid data after data cleaning is vectorized according to the first processing instruction, and high-impact features are screened out, which lays a solid foundation for subsequent accurate data judgment.

[0059] Furthermore, the embodiment of the present application also includes:

[0060] Step S650: obtaining first influencing factor information according to the first bank device information, wherein the first influencing factor information has a first degree of influence with the first bank device;

[0061] Step S660: obtaining a second processing instruction;

[0062] Step S670: According to the second processing instruction, the missing values ​​of the first valid data after the vectorization processing are supplemented to supplement the missing values ​​of the first influencing factor information.

[0063] Specifically, based on the first bank equipment information, influencing factor information having a first degree of influence on the first equipment is obtained, and based on the influencing factor information, a second processing instruction is obtained. Based on the second processing instruction, the first valid data is judged to determine whether there is a missing influencing factor in the first valid data. If so, the first valid data is supplemented with the missing value to supplement the missing value of the first influencing factor information, so as to obtain more complete first valid data, thereby laying a foundation for further judgment on the data.

[0064] Furthermore, the embodiment of the present application also includes:

[0065] Step S671: obtaining the second influencing factor information in the first valid data after the missing value is supplemented;

[0066] Step S672: Obtaining first preset impact threshold information;

[0067] Step S673: Obtain a first deletion instruction;

[0068] Step S674: according to the first deletion instruction, deleting the influencing factors in the second influencing factor information that do not meet the first preset influencing threshold information;

[0069] Step S675: obtaining second preset impact threshold information;

[0070] Step S676: Obtain a first screening instruction;

[0071] Step S677: According to the first screening instruction, a complete search, a heuristic search, or a random algorithm is used to screen the influencing factors that meet the second preset influencing threshold information in the second influencing factor information after the influencing factors are deleted to obtain the first feature information.

[0072] Specifically, the influence of the influencing factor is determined based on the first preset influence threshold information. Specifically, the first preset influence threshold information is set first, and the influencing factors that do not meet the threshold are deleted based on the first preset influence threshold information, that is, the feature process of deleting small influencing factors. According to the influencing factors that have been deleted, the second preset influence threshold information is obtained. According to the second preset influence threshold information, the first screening instruction is obtained. According to the first screening instruction, the influencing factors that meet the second preset influence threshold information in the second influencing factor information after the influencing factors are deleted are screened to obtain the first feature information by using a complete search, a heuristic search, and a random algorithm, thereby screening out high-influence features through the three methods of a complete search, a heuristic search, and a random algorithm, so as to prepare for the next step of classification. Through the above setting of the influencing factor threshold, the influencing factors are deleted and screened based on the first preset influence threshold information and the second preset influence threshold information, so as to obtain more accurate features with greater influence on the data mutation, laying the foundation for subsequent data analysis.

[0073] Furthermore, the first training model is adjusted according to the first verification result and the first prediction classification result. Step S1100 of the embodiment of the present application further includes:

[0074] Step S1110: obtaining a first accuracy and a first recall of the first verification result according to the first verification result and the first prediction classification result;

[0075] Step S1120: Calculate an F1-Score value according to the first precision and the first recall;

[0076] Step S1130: obtaining preset accuracy threshold information;

[0077] Step S1140: Determine whether the F1-Score value meets the preset accuracy threshold information;

[0078] Step S1150: If the F1-Score value satisfies the preset accuracy threshold information, the first original data information is added to the first training model to adjust the first training model.

[0079] Furthermore, the embodiment of the present application also includes:

[0080] Step S1160: If the F1-Score value does not meet the preset accuracy threshold information, obtaining first deviation information;

[0081] Step S1170: obtaining a first correction instruction according to the first deviation information;

[0082] Step S1180: adjusting and training the first training model according to the first correction instruction and the first deviation information.

[0083] Specifically, based on the first verification result and the first prediction classification result, the first accuracy and the first recall of the first verification result are obtained. Based on the first accuracy and the first recall, the F1-Score is finally calculated to represent the comprehensive accuracy. The trained model is retrained (fine-tuned) according to the F1-Score result. In detail, the preset accuracy threshold information is obtained, and it is determined whether the F1-Score value meets the preset accuracy threshold information. When the F1-Score value meets the preset accuracy threshold information, it indicates that the training model obtained by training the SVM and naive Bayes algorithm based on the first sample data is accurate when processing the first original data. At this time, the first original data information is used as sample data to continue training the first training model, so that the first training model can perform more accurate analysis and judgment; if the F1-Score value does not meet the preset accuracy threshold information, the first deviation information is obtained, and the first correction instruction is obtained according to the first deviation information. According to the first correction instruction, the first training model is adjusted to ensure the accuracy of the first training model in processing the input data, thereby achieving active and accurate early warning of possible anomalies, reducing the impact of faults on the system, and achieving the technical effect of improving operation and maintenance efficiency.

[0084] In summary, the bank equipment failure early warning method and system based on machine learning provided by the embodiment of the present application has the following technical effects:

[0085] 1. Due to the use of SVM and naive Bayes algorithm to train the first training model based on the first sample data, the first original data is input into the first training model to obtain the output information of the first training model, the first original data is passed through the feature engineering unit to obtain the second classification result information, and it is determined whether to perform manual investigation according to the second classification result information, and the first training model is adjusted according to the manual investigation result, so as to obtain a more accurate first training model to process the input data, so as to achieve active and accurate early warning of possible anomalies, reduce the impact of faults on the system, and achieve the technical effect of improving operation and maintenance efficiency.

[0086] 2. Since the method of data cleaning is adopted for the first data, the technical effect of ensuring the integrity and reliability of the data is achieved, the first valid data after data cleaning is vectorized according to the first processing instruction, and high-impact features are screened out, which lays a solid foundation for subsequent accurate data judgment.

[0087] 3. By adjusting the first training model based on the F1-Score value, the accuracy of the first training model in processing the input data is guaranteed, thereby achieving active and accurate early warning of possible anomalies, reducing the impact of failures on the system, and achieving the technical effect of improving operation and maintenance efficiency.

[0088] Embodiment 2

[0089] Based on the same inventive concept as the bank equipment failure early warning method based on machine learning in the aforementioned embodiment, the present invention also provides a bank equipment failure early warning system based on machine learning, such as Figure 2 As shown, the system comprises:

[0090] A first obtaining unit 11, the first obtaining unit 11 is used to obtain first training model information, wherein the first training model is a binary classification prediction model;

[0091] A second obtaining unit 12, the second obtaining unit 12 is used to obtain first sample data from the first bank device information, wherein the first sample data includes negative sample data and positive sample data;

[0092] A first training unit 13, wherein the first training unit 13 is used to train the first training model using SVM and naive Bayes algorithm according to the first sample data;

[0093] A third obtaining unit 14, wherein the third obtaining unit 14 is used to obtain first original data information through the original data monitoring unit;

[0094] A first input unit 15, wherein the first input unit 15 is used to input the first original data information into the first training model after training, to obtain output information of the first training model, wherein the output information is first prediction classification result information;

[0095] A fourth obtaining unit 16, the fourth obtaining unit 16 is used to obtain first feature information through a feature engineering unit according to the first original data information;

[0096] A fifth obtaining unit 17, wherein the fifth obtaining unit 17 is used to obtain second classification result information according to the first feature information;

[0097] a first determining unit 18, wherein the first determining unit 18 is used to determine whether to issue a first warning instruction according to the second classification result information;

[0098] A sixth obtaining unit 19, wherein the sixth obtaining unit 19 is used to send the first warning instruction to the manual checking unit through the alarm interface if a first warning instruction needs to be issued, so that the administrator performs manual checking after receiving the first warning instruction;

[0099] A seventh obtaining unit 20, the seventh obtaining unit 20 is used to obtain the first verification result of the manual investigation;

[0100] The first adjustment unit 21 is used to adjust the first training model according to the first verification result and the first prediction classification result.

[0101] Furthermore, the system also includes:

[0102] an eighth obtaining unit, the eighth obtaining unit being used to obtain a first preset time period;

[0103] a ninth obtaining unit, configured to obtain first abnormal event information according to the first preset time period;

[0104] a tenth obtaining unit, configured to obtain a second preset time period according to the first abnormal event information, wherein the second preset time period is before the first abnormal event occurs;

[0105] an eleventh obtaining unit, the eleventh obtaining unit being configured to obtain the negative sample data according to the second preset time period;

[0106] a twelfth obtaining unit, the twelfth obtaining unit being used to obtain a third preset time period;

[0107] a thirteenth obtaining unit, the thirteenth obtaining unit being configured to obtain the positive sample data according to the third preset time period;

[0108] A fourteenth obtaining unit, the fourteenth obtaining unit is used to obtain the first sample data according to the negative sample data and the positive sample data.

[0109] Furthermore, the system also includes:

[0110] a fifteenth obtaining unit, the fifteenth obtaining unit being used to obtain a first cleaning instruction;

[0111] a sixteenth obtaining unit, configured to obtain first valid data after cleaning the first original data according to the first cleaning instruction;

[0112] a seventeenth obtaining unit, the seventeenth obtaining unit being used to obtain a first processing instruction;

[0113] A first processing unit, wherein the first processing unit is used to perform vectorized processing on the first valid data according to the first processing instruction.

[0114] Furthermore, the system also includes:

[0115] an eighteenth obtaining unit, the eighteenth obtaining unit being configured to obtain first influencing factor information according to the first bank device information, wherein the first influencing factor information has a first degree of influence with the first bank device;

[0116] a nineteenth obtaining unit, the nineteenth obtaining unit being used to obtain a second processing instruction;

[0117] A first supplementing unit, wherein the first supplementing unit is used to supplement missing values ​​of the first valid data after the vectorization processing according to the second processing instruction, so as to supplement the missing values ​​of the first influencing factor information.

[0118] Furthermore, the system also includes:

[0119] a twentieth obtaining unit, the twentieth obtaining unit being used to obtain second influencing factor information in the first valid data after the missing value is supplemented;

[0120] A twenty-first obtaining unit, the twenty-first obtaining unit is used to obtain first preset impact threshold information;

[0121] A twenty-second obtaining unit, the twenty-second obtaining unit is used to obtain a first deletion instruction;

[0122] a first deleting unit, configured to delete, according to the first deleting instruction, the influencing factors in the second influencing factor information that do not meet the first preset influencing threshold information;

[0123] A twenty-third obtaining unit, the twenty-third obtaining unit is used to obtain second preset impact threshold information;

[0124] a twenty-fourth obtaining unit, the twenty-fourth obtaining unit being used to obtain a first screening instruction;

[0125] The twenty-fifth obtaining unit is used to screen the influencing factors that meet the second preset influencing threshold information in the second influencing factor information after the influencing factors are deleted according to the first screening instruction, using a complete search, a heuristic search, or a random algorithm to obtain the first feature information.

[0126] Furthermore, the system also includes:

[0127] a twenty-sixth obtaining unit, the twenty-sixth obtaining unit being configured to obtain a first accuracy and a first recall rate of the first verification result according to the first verification result and the first prediction classification result;

[0128] a twenty-seventh obtaining unit, wherein the twenty-seventh obtaining unit is used to calculate an F1-Score value according to the first precision rate and the first recall rate;

[0129] A twenty-eighth obtaining unit, the twenty-eighth obtaining unit is used to obtain preset accuracy threshold information;

[0130] A first judgment unit, the first judgment unit is used to judge whether the F1-Score value meets the preset accuracy threshold information;

[0131] A second adjustment unit, wherein the second adjustment unit is configured to add the first original data information to the first training model to adjust the first training model if the F1-Score value meets the preset accuracy threshold information.

[0132] Furthermore, the system also includes:

[0133] a twenty-ninth obtaining unit, the twenty-ninth obtaining unit being configured to obtain first deviation information if the F1-Score value does not satisfy the preset accuracy threshold information;

[0134] a thirtieth obtaining unit, the thirtieth obtaining unit being configured to obtain a first correction instruction according to the first deviation information;

[0135] A third adjustment unit, wherein the third adjustment unit is used to adjust and train the first training model according to the first correction instruction and the first deviation information.

[0136] The foregoing Figure 1 The various variations and specific examples of a bank equipment failure warning method based on machine learning in Example 1 are also applicable to a bank equipment failure warning system based on machine learning in this embodiment. Through the above detailed description of a bank equipment failure warning method based on machine learning, those skilled in the art can clearly know the implementation method of a bank equipment failure warning system based on machine learning in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here.

[0137] Exemplary Electronic Devices

[0138] Reference below Figure 3 To describe the electronic device of the embodiment of the present application.

[0139] Figure 3The figure shows a schematic structural diagram of an electronic device according to an embodiment of the present application.

[0140] Based on the inventive concept of a bank equipment failure warning method based on machine learning in the aforementioned embodiment, the present invention also provides a bank equipment failure warning system based on machine learning, on which a computer program is stored, and when the program is executed by a processor, the steps of any method of the bank equipment failure warning method based on machine learning described above are implemented.

[0141] Among them, Figure 3 In the embodiment of the present invention, a bus architecture (represented by bus 300) is shown, which may include any number of interconnected buses and bridges, and bus 300 links together various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. Bus interface 306 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, namely a transceiver, which provides a unit for communicating with various other systems over a transmission medium.

[0142] The processor 302 is responsible for managing the bus 300 and general processing, while the memory 304 may be used to store data used by the processor 302 when performing operations.

[0143] An embodiment of the present invention provides a bank equipment failure early warning method based on machine learning, the method is applied to a risk assessment system, and the system includes a model training unit, a raw data monitoring unit, a feature engineering unit, a classification algorithm unit, an early warning unit, and a manual verification unit, wherein the method includes: obtaining first training model information, wherein the first training model is a binary classification prediction model; obtaining first sample data from first bank equipment information, wherein the first sample data includes negative sample data and positive sample data; according to the first sample data, using SVM and naive Bayes algorithm to train the first training model; obtaining first raw data information through the raw data monitoring unit; inputting the first raw data information into the system; into the first training model after training, and obtain the output information of the first training model, wherein the output information is the first prediction classification result information; according to the first original data information, obtain the first feature information through the feature engineering unit; according to the first feature information, obtain the second classification result information; according to the second classification result information, determine whether to issue a first warning instruction; if it is necessary to issue a first warning instruction, send the first warning instruction to the manual verification unit through the alarm interface, so that the administrator can perform manual troubleshooting after receiving the first warning instruction; obtain the first verification result of the manual troubleshooting; according to the first verification result and the first prediction classification result, adjust the first training model. The technical problem that the bank cannot or cannot accurately provide active warning for possible abnormalities in the prior art, thereby affecting the operation and maintenance efficiency, is solved, and the possible abnormalities can be actively and accurately warned, the impact of failures on the system is reduced, and the technical effect of improving the operation and maintenance efficiency is achieved.

[0144] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0145] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or multiple boxes.

[0146] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0147] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in a box or multiple boxes. Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the attached claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0148] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A bank equipment failure early warning method based on machine learning, the method is applied to a risk assessment system, and the system includes a model training unit, a raw data monitoring unit, a feature engineering unit, a classification algorithm unit, an early warning unit, and a manual verification unit, wherein: The method comprises: Obtaining first training model information, wherein the first training model is a binary classification prediction model; Obtaining first sample data from first bank device information, wherein the first sample data includes negative sample data and positive sample data; According to the first sample data, the first training model is trained using SVM and naive Bayes algorithm; Obtaining first original data information through the original data monitoring unit; Inputting the first original data information into the first training model after training is completed to obtain output information of the first training model, wherein the output information is first prediction classification result information; According to the first original data information, obtaining first feature information through the feature engineering unit; The feature engineering unit is a unit for performing feature comparison and screening; Obtaining second classification result information according to the first feature information; Determining whether to issue a first warning instruction according to the second classification result information; If a first warning instruction needs to be issued, the first warning instruction is sent to the manual checking unit through the alarm interface, so that the administrator performs manual checking after receiving the first warning instruction; Obtaining a first verification result of the manual investigation; The first training model is adjusted according to the first verification result and the first prediction classification result.

2. The method of claim 1, wherein: The obtaining of first sample data from first bank device information includes: Obtaining a first preset time period; Obtaining first abnormal event information according to the first preset time period; According to the first abnormal event information, a second preset time period is obtained, wherein the second preset time period is before the first abnormal event occurs; According to the second preset time period, obtaining the negative sample data; Obtaining a third preset time period; According to the third preset time period, obtaining the positive sample data; The first sample data is obtained according to the negative sample data and the positive sample data.

3. The method of claim 1, wherein: The step of obtaining first feature information by the feature engineering unit according to the first original data information includes: Obtaining the first cleaning instruction; After cleaning the first original data according to the first cleaning instruction, first valid data is obtained; Obtaining a first processing instruction; According to the first processing instruction, vectorized processing is performed on the first valid data.

4. The method of claim 3, wherein: The method further comprises: Obtaining first impact factor information according to the first bank device information, wherein the first impact factor information has a first impact degree with the first bank device; Obtaining a second processing instruction; According to the second processing instruction, missing values ​​are supplemented for the first valid data after the vectorization processing to supplement the missing values ​​of the first influencing factor information.

5. The method of claim 4, wherein: The method further comprises: Obtaining second influencing factor information in the first valid data after the missing value is supplemented; Obtaining first preset impact threshold information; Get the first deletion instruction; Deleting, according to the first deletion instruction, the influencing factors in the second influencing factor information that do not meet the first preset influencing threshold information; Obtaining second preset impact threshold information; Obtaining a first screening instruction; According to the first screening instruction, a complete search, a heuristic search, or a random algorithm is used to screen the influencing factors that meet the second preset influencing threshold information in the second influencing factor information after deleting the influencing factors, so as to obtain the first feature information.

6. The method of claim 1, wherein: The adjusting the first training model according to the first verification result and the first prediction classification result includes: Obtaining a first accuracy rate and a first recall rate of the first verification result according to the first verification result and the first prediction classification result; Calculate the F1-Score value according to the first accuracy and the first recall; Obtaining preset accuracy threshold information; Determine whether the F1-Score value meets the preset accuracy threshold information; If the F1-Score value meets the preset accuracy threshold information, the first original data information is added to the first training model, and the first training model is adjusted.

7. The method of claim 6, wherein: The method further comprises: If the F1-Score value does not meet the preset accuracy threshold information, obtaining first deviation information; Obtaining a first correction instruction according to the first deviation information; The first training model is adjusted and trained according to the first correction instruction and the first deviation information.

8. A bank equipment failure early warning system based on machine learning, wherein: The system comprises: A first obtaining unit, the first obtaining unit is used to obtain first training model information, wherein the first training model is a binary classification prediction model; a second obtaining unit, the second obtaining unit being used to obtain first sample data from the first bank device information, wherein the first sample data includes negative sample data and positive sample data; A first training unit, wherein the first training unit is used to train the first training model using SVM and naive Bayes algorithm according to the first sample data; A third obtaining unit, the third obtaining unit is used to obtain the first original data information through the original data monitoring unit; A first input unit, wherein the first input unit is used to input the first original data information into the first training model after training, to obtain output information of the first training model, wherein the output information is first prediction classification result information; A fourth obtaining unit, the fourth obtaining unit is used to obtain first feature information through a feature engineering unit according to the first original data information, wherein the feature engineering unit is a unit for performing feature comparison and screening; a fifth obtaining unit, configured to obtain second classification result information according to the first feature information; a first determining unit, the first determining unit being used to determine whether to issue a first warning instruction according to the second classification result information; a sixth obtaining unit, wherein the sixth obtaining unit is used to send the first warning instruction to the manual checking unit through the alarm interface if a first warning instruction needs to be issued, so that the administrator performs manual checking after receiving the first warning instruction; A seventh obtaining unit, the seventh obtaining unit is used to obtain the first verification result of the manual investigation; A first adjustment unit, wherein the first adjustment unit is used to adjust the first training model according to the first verification result and the first prediction classification result.

9. A bank equipment failure early warning system based on machine learning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images