Training method, system, device and medium for solution path prediction model

Abstract

The present invention discloses a training method, system, device and medium for a solution prediction model, the training method includes obtaining a verification data set, inputting the verification data set into an initial solution prediction model for training to obtain the predicted solution corresponding to each event; screening abnormal events whose predicted solution does not match the actual solution; calculating the confidence of the abnormal event based on N standard solution paths and the predicted solution paths of the abnormal event; screening a number of abnormal events whose confidence meets a preset threshold range, and correcting the actual solution paths of the several abnormal events; training the initial solution prediction model based on the corrected verification data set to obtain an optimized solution prediction model. The present invention improves the data accuracy of the verification data set, realizes the optimization of the solution prediction model, improves the accuracy of the predicted solution, reduces the workload of manually correcting the verification data set, and improves work efficiency.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a training method, system, device and medium for a solution path prediction model. Background Art

[0002] In many large and medium-sized enterprises, as the company's business develops and the operation time continues, more and more historical operation data is generated. Through machine learning training of historical operation data, the strategy automation of certain IT (Information Technology) links can be realized. The most important thing for machine learning models is the accuracy and consistency of input sample data. However, in the actual operation process, there are various abnormal scenarios and human factors, which lead to errors in the input sample data archived in the operation system.

[0003] In traditional machine learning models, the data in the input sample set needs to be manually checked and proofread one by one, and the errors and special data need to be eliminated and corrected. However, due to the huge amount of historical data, manual labeling will consume a lot of manpower. In addition, during the process of operational iteration, the system will continuously generate new data, which will lead to a long-term large amount of manpower consumption, increase the workload of employees, and is not conducive to improving work efficiency. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the defects in the traditional machine learning model in the prior art, that is, the input sample data of the model is inaccurate, and manual inspection, proofreading, elimination and correction of the data in the input sample data are required one by one, and the manual inspection of data is huge in workload and is not conducive to improving work efficiency. A solution to the problem of training a prediction model is proposed, including a method, system, equipment and medium.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] In a first aspect, a training method for a solution pathway prediction model is provided, the training method comprising:

[0007] Acquire a verification data set, the verification data set including a plurality of events, N standard solutions and an actual solution corresponding to each of the events; the actual solution is one of the N standard solutions;

[0008] Inputting the verification data set into the initial solution path prediction model for training to obtain the predicted solution paths corresponding to each of the events; the initial solution path prediction model takes the events as input and takes the predicted solution paths corresponding to the events as output;

[0009] Screen out abnormal events where the predicted solution path does not match the actual solution path;

[0010] Calculating the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event;

[0011] Screening a number of abnormal events whose confidences meet a preset threshold range, and revising actual solutions to the number of abnormal events;

[0012] The initial solution pathway prediction model is trained based on the revised validation data set to obtain an optimized solution pathway prediction model.

[0013] The training method of the solution path prediction model of the present invention obtains a verification data set, inputs the verification data set into the initial solution path prediction model for training to obtain the predicted solution path corresponding to each event; screens abnormal events whose predicted solution paths are inconsistent with the actual solution paths; calculates the confidence of the abnormal event based on N standard solution paths and the predicted solution paths of the abnormal event; screens a number of abnormal events whose confidence paths meet a preset threshold range, and corrects the actual solution paths of the several abnormal events; trains the initial solution path prediction model based on the corrected verification data set to obtain an optimized solution path prediction model; improves the data accuracy of the verification data set, realizes the optimization of the solution path prediction model, improves the accuracy of the predicted solution path, reduces the workload of manually correcting the verification data set, and improves work efficiency.

[0014] Preferably, the step of calculating the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event specifically includes:

[0015] The distance values ​​between the predicted solution vector corresponding to the predicted solution and the N standard solution vectors corresponding to the N standard solutions are respectively calculated; and the confidence is calculated based on the distance values.

[0016] Preferably, the distance value includes a minimum distance Dmin and N-1 remaining distances Drest excluding the minimum distance Dmin;

[0017] Calculate the average distance Dave of the N-1 remaining distances Drest;

[0018] The step of calculating the confidence level based on the distance value specifically includes:

[0019] According to the formula The confidence level is calculated, where T is the confidence level.

[0020] Preferably, the preset threshold range includes a first preset threshold range and a second preset threshold range; an upper limit value of the second preset threshold range is greater than an upper limit value of the first preset threshold range;

[0021] The step of screening a number of abnormal events whose confidences meet a preset threshold range and correcting actual solutions to the number of abnormal events specifically includes:

[0022] Screening a number of abnormal events whose confidences meet the first preset threshold range, and revising actual solutions to the number of abnormal events;

[0023] Training the initial solution pathway prediction model based on the revised validation data set to obtain an optimized solution pathway prediction model;

[0024] Return to the step of screening abnormal events whose predicted solution paths do not match the actual solution paths, until the confidence level meets the second preset threshold range.

[0025] The training method of the solution prediction model of the present invention includes a preset threshold range including a first preset threshold range and a second preset threshold range, and a plurality of abnormal events whose confidences meet the first preset threshold range are screened, and the actual solutions of the plurality of abnormal events are corrected; the initial solution prediction model is trained based on the corrected verification data set to obtain an optimized solution prediction model; and the step of screening abnormal events whose predicted solution does not match the actual solution is returned until the confidence meets the second preset threshold range. The solution prediction model is continuously optimized to improve the accuracy of the predicted solution.

[0026] In a second aspect, a solution path prediction system is provided, the solution path prediction system comprising: a verification data set acquisition module, used to acquire a verification data set, the verification data set comprising multiple

[0027] An event, N standard solutions and an actual solution corresponding to each of the events; the actual solution being one of the N standard solutions;

[0028] A prediction solution path training module, used for inputting the verification data set into an initial solution path prediction model for training to obtain a prediction solution path corresponding to each of the events; the initial solution path prediction model takes the event as input and takes the prediction solution path corresponding to the event as output;

[0029] The first screening module is used to screen abnormal events where the predicted solution path does not match the actual solution path;

[0030] A confidence calculation module, used to calculate the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event;

[0031] A second screening module is used to screen a number of abnormal events whose confidences meet a preset threshold range, and to modify actual solutions to the number of abnormal events;

[0032] The model training module is used to train the initial solution path prediction model based on the revised verification data set to obtain an optimized solution path prediction model.

[0033] Preferably, the confidence calculation module includes:

[0034] A distance value calculation unit is used to respectively calculate the distance values ​​between the predicted solution vector corresponding to the predicted solution and the N standard solution vectors corresponding to the N standard solutions; and calculate the confidence based on the distance value.

[0035] Preferably, the distance value includes a minimum distance Dmin and N-1 remaining distances Drest excluding the minimum distance Dmin;

[0036] The distance value calculation unit is specifically used to calculate the average distance Dave of the N-1 remaining distances Drest; and according to the formula The confidence level is calculated, where T is the confidence level.

[0037] Preferably, the preset threshold range includes a first preset threshold range and a second preset threshold range; an upper limit value of the second preset threshold range is greater than an upper limit value of the first preset threshold range;

[0038] The second screening module is specifically used to screen a number of abnormal events whose confidences meet the first preset threshold range, and to modify actual solutions to the number of abnormal events;

[0039] The model training module is specifically used to train the initial solution path prediction model based on the revised verification data set to obtain an optimized solution path prediction model;

[0040] The first screening module is called again to screen abnormal events for which the predicted solution path does not match the actual solution path, until the confidence level meets the second preset threshold range.

[0041] The solution prediction system of the present invention includes a verification data set acquisition module, a prediction solution training module, a first screening module, a confidence calculation module, a second screening module and a model training module. Through the mutual cooperation of each module, the data accuracy of the verification data set in the solution prediction system is improved, the solution prediction model is optimized, the accuracy of the predicted solution is improved, the workload of manually correcting the verification data set is reduced, and the work efficiency is improved.

[0042] According to a third aspect, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the training method of any of the solution path prediction models described above is implemented.

[0043] In a fourth aspect, a computer-readable medium is provided, on which computer instructions are stored, and when the computer instructions are executed by a processor, the training method of any of the solution path prediction models described above is implemented.

[0044] The positive and progressive effects of the present invention are:

[0045] The training method of the solution prediction model of the present invention obtains a verification data set, the verification data set includes multiple events, N standard solutions and actual solutions corresponding to each event; the actual solution is one of the N standard solutions; the verification data set is input into the initial solution prediction model for training to obtain the predicted solution corresponding to each event; the initial solution prediction model takes the event as input and the predicted solution corresponding to the event as output; screens abnormal events whose predicted solution is inconsistent with the actual solution; calculates the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event; screens several abnormal events whose confidence meets the preset threshold range, and corrects the actual solution of several abnormal events; trains the initial solution prediction model based on the corrected verification data set to obtain an optimized solution prediction model. The present invention improves the data accuracy of the verification data set, realizes the optimization of the solution prediction model, improves the accuracy of the predicted solution, reduces the workload of manually correcting the verification data set, and improves work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic diagram of a flow chart of a method for training a solution path prediction model provided in Example 1 of the present invention;

[0047] Figure 2 A schematic flow chart of another method for training a solution prediction model provided in Example 1 of the present invention;

[0048] Figure 3A schematic diagram of a solution path prediction system provided in Example 2 of the present invention;

[0049] Figure 4 This is a schematic diagram of the structure of an electronic device shown in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0050] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0051] Example 1

[0052] This embodiment provides a training method for solving a pathway prediction model. Figure 1 A flowchart of a training method for a solution prediction model provided in this embodiment is shown as follows: Figure 1 As shown, the training method of the solution pathway prediction model includes:

[0053] Step 101: Obtain a verification data set.

[0054] The verification data set includes multiple events, N standard solutions and actual solutions corresponding to each event; the actual solution is one of the N standard solutions.

[0055] In this step, each event corresponds to an actual solution, and the actual solution is one of the N standard solutions in the validation data set in the sample set, where N can be an integer greater than or equal to 2.

[0056] Step 102: Input the verification data set into the initial solution path prediction model for training to obtain the predicted solution path corresponding to each event.

[0057] Among them, the initial solution path prediction model takes events as input and the predicted solution path corresponding to the events as output.

[0058] In this step, the solution pathways of the events in the validation data set are predicted based on the solution pathway prediction model to obtain predicted solution pathways. The validation data set is the input sample set of the initial solution pathway prediction model.

[0059] Step 103: Screen abnormal events whose predicted solutions do not match the actual solutions.

[0060] In this step, the predicted solution path and the actual solution path of each event are marked to screen out abnormal events whose predicted solution path does not match the actual solution path.

[0061] Step 104: Calculate the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event.

[0062] Among them, confidence is used to characterize the degree of closeness between the predicted solution path of an event and the actual solution path, and is used to characterize the degree of confidence of the solution path prediction model in the accuracy of the predicted solution path.

[0063] Step 105: Screen a number of abnormal events whose confidences meet a preset threshold range, and modify actual solutions to the abnormal events.

[0064] In this step, the confidences corresponding to the events are sorted, several abnormal events whose confidences meet the preset threshold range are screened out, and the actual solutions to the abnormal events are corrected, that is, the verification data set is corrected to ensure the accuracy of the actual solutions to the abnormal events.

[0065] Step 106: training the initial solution path prediction model based on the revised validation data set to obtain an optimized solution path prediction model.

[0066] In this step, the initial solution pathway prediction model is trained using the corrected validation data set to obtain an optimized solution pathway prediction model.

[0067] The training method of the solution path prediction model of the present embodiment obtains a verification data set, inputs the verification data set into the initial solution path prediction model for training to obtain the predicted solution path corresponding to each event; screens abnormal events whose predicted solution paths do not match the actual solution paths; calculates the confidence of the abnormal event based on N standard solution paths and the predicted solution paths of the abnormal event; screens several abnormal events whose confidence paths meet a preset threshold range, and corrects the actual solution paths of the several abnormal events; trains the initial solution path prediction model based on the corrected verification data set to obtain an optimized solution path prediction model; improves the data accuracy of the verification data set, realizes the optimization of the solution path prediction model, improves the accuracy of the predicted solution path, reduces the workload of manually correcting the verification data set, and improves work efficiency.

[0068] In an optional implementation, the above step 104 specifically includes: respectively calculating the distance values ​​between the predicted solution vector corresponding to the predicted solution and the N standard solution vectors corresponding to the N standard solutions; and calculating the confidence based on the distance values.

[0069] In this embodiment, the predicted solution corresponds to a predicted solution vector, the standard solution corresponds to a standard solution vector, the distance value between the predicted solution vector and N standard solution vectors is calculated, and the confidence is calculated based on the distance value.

[0070] In an optional embodiment, the distance value includes the minimum distance Dmin and N-1 remaining distances Drest excluding the minimum distance Dmin; the step of calculating the confidence based on the distance value specifically includes: calculating the average distance value Dave of the N-1 remaining distances Drest; according to the formula Calculate the confidence level, where T is the confidence level.

[0071] In an alternative embodiment, Figure 2 A flowchart of a training method for a solution path prediction model provided in this embodiment, wherein the preset threshold range includes a first preset threshold range and a second preset threshold range; the upper limit value of the second preset threshold range is greater than the upper limit value of the first preset threshold range; Figure 2 As shown, the above step 104 further includes:

[0072] Step 107: Determine whether the confidence level meets a second preset threshold range.

[0073] If not, step 105 is executed until the confidence level meets the second preset threshold range. The above step 105 specifically includes: screening a number of abnormal events whose confidence levels meet the first preset threshold range, and modifying the actual solutions to the abnormal events.

[0074] If yes, the training ends.

[0075] In this embodiment, the upper limit value of the second preset threshold range is greater than the upper limit value of the first preset threshold range. The actual solutions to several abnormal events whose confidence levels satisfy the first preset threshold range are first corrected, and the corrected verification data set is obtained to train the initial solution prediction model to obtain the optimized solution prediction model. The screening is repeated multiple times to ensure that the confidence levels of the abnormal events satisfy the preset second threshold range, thereby continuously optimizing the solution prediction model to improve the accuracy of predicting solution paths.

[0076] The training method of the solution prediction model of this embodiment, the preset threshold range includes a first preset threshold range and a second preset threshold range, the upper limit value of the second preset threshold range is greater than the upper limit value of the first preset threshold range, by screening a number of abnormal events whose confidences meet the first preset threshold range, and correcting the actual solutions of the several abnormal events; training the initial solution prediction model based on the corrected verification data set to obtain an optimized solution prediction model; returning to the step of screening abnormal events whose predicted solution does not match the actual solution, until the confidence meets the second preset threshold range. The continuous optimization of the solution prediction model is achieved, and the accuracy of predicting the solution is improved.

[0077] In an optional implementation, text encoding is performed on the verification data set and feature values ​​are extracted, and the feature values ​​are input into the initial solution path prediction model for training to obtain the predicted solution path corresponding to each event.

[0078] The training method of the solution prediction model of the present invention is described in detail below in conjunction with specific application scenarios.

[0079] Export all event reports for the past year in batches from the company's event management system. For example, select 2,000 events, and select 100 events from these 2,000 events as the training data set. The remaining 1,900 events are the verification data set. Check the actual solution path corresponding to each event in the training data set one by one to ensure that the actual solution path corresponding to each event is correct. For example, if the event is a computer failure, network failure, or phone failure, the actual solution path corresponding to the computer failure is the desktop team, the solution path corresponding to the network failure is the network team, and the actual solution path corresponding to the phone failure is the phone team.

[0080] For example, the title corresponding to the event computer failure is computer failure, and the symptom is computer blue screen. Use the generated initial solution path prediction model to encode and predict the 1,900 events in the verification data set to generate a predicted solution path. Since the actual solution paths for each event in the verification data set have not been checked one by one, there may be errors in the actual solution paths. The abnormal events in the verification data set whose predicted solution paths do not match the actual solution paths are screened out; there are N standard solutions in the verification data set, for example, there are 3 annotated solutions, then the confidence of the abnormal event is calculated based on the 3 standard solutions and the predicted solution paths of the abnormal event; screen several abnormal events whose confidence paths meet the preset threshold range, and correct the actual solution paths of several abnormal events; train the initial solution path prediction model based on the corrected verification data set to obtain an optimized solution path prediction model.

[0081] For example, if the number of abnormal events is 100, there are 100 corresponding confidence outputs. The obtained confidences are sorted from high to low according to the confidence values ​​to obtain a confidence sorting table. The first preset threshold range is set to 0.50 to 0.80, and the second preset threshold range is set to 0.90 to 0.95, wherein the upper limit value of the second preset threshold range, 0.95, is greater than the upper limit value of the first preset threshold range, 0.80. The confidence of the abnormal event is less than 1 and infinitely close to 1. It is determined whether the confidence of the 100 abnormal events meets the second preset threshold range. If so, the training is terminated. If not, several abnormal events whose confidence meets the first preset threshold range of 0.50 to 0.80 are screened, and the corresponding actual solutions are corrected. The initial solution path prediction model is trained based on the revised validation data set to obtain an optimized solution path prediction model; abnormal events whose predicted solution paths are inconsistent with the actual solution paths are continued to be screened. If the number of abnormal events is 40 at this time, it is determined whether the confidence of the 40 abnormal events meets the second preset threshold range of 0.90 to 0.95. If so, the training is terminated; if not, several abnormal events whose confidence meets the first preset threshold range of 0.50 to 0.80 are screened, and the corresponding actual solution paths are corrected. The initial solution path prediction model is trained based on the corrected validation data set to obtain an optimized solution path prediction model; abnormal events whose predicted solution paths are inconsistent with the actual solution paths are continued to be screened. If the number of abnormal events is 10 at this time, it is determined whether the confidence of the 10 abnormal events meets the second preset threshold range of 0.90 to 0.95. If so, the training is terminated; if not, the actual solution paths of the 10 abnormal events are continued to be corrected. After multiple rounds of correction and training, the number of abnormal events will become smaller and smaller until the confidence of the abnormal events meets the second preset threshold range. Then the correction is stopped, and the final optimized solution path prediction model is obtained.

[0082] In this application scenario, the first preset threshold range is set to 0.50 to 0.80, and the second preset threshold range is set to 0.90 to 0.95, but the first preset threshold range and the second preset threshold range are not limited to this. The first preset threshold range can also be 0.45 to 0.85, and the second preset threshold range can be set to 0.80 to 0.98, as long as the upper limit value of the second preset threshold range, such as 0.98, is greater than the upper limit value of the first preset threshold range, such as 0.8.

[0083] The "title" and "symptoms" of the events in the training set can be encoded with One Hot Encoder and TF-IDF (term frequency–inverse document frequency, a commonly used weighting technique for information retrieval and data mining) text to extract feature values, and then the PyTorch (an open source Python machine learning library for applications such as natural language processing) framework is used to train and generate an initial solution path prediction model. In classification and clustering operations, we often calculate the distance between two individuals. This is not a problem for continuous numbers, but for nominal categories, it is difficult to calculate the distance. Even if the categories are mapped to numbers, such as (A, B, C) and (0, 1, 2), we cannot assume that the distance between A and B is 0, the distance between B and C is 1, and the distance between A and C is 2. The one-hot encoding method is designed to deal with this distance measurement, and this method assumes that the distance between each category is the same. This method maps categories to vectors, for example (A, B, C) maps to (1, 0, 0), (0, 1, 0), (0, 0, 1) respectively. Note that the Euclidean distances between categories are now the same, and the 1 in the vector only represents the category, but the data is in numerical format.

[0084] For example, the predicted solution path vector corresponding to the predicted solution path of a certain abnormal event is (X, Y, Z), and the three standard solution path vectors corresponding to the three labeled solution paths are (X N , Y N , Z N ), where N is 3, i.e. (X1, Y1, Z1), (X2, Y2, Z2) and (X3, Y3, Z3), then the distance values ​​D between the vector (X, Y, Z) and the vectors (X1, Y1, Z1), (X2, Y2, Z2) and (X3, Y3, Z3) are calculated respectively. N , D N =(X N -X) 2 +(Y N -Y) 2 +(Z N -Z) 2 , get the distance values ​​D1, D2, and D3. If the value of D1 is the smallest, that is, the minimum distance Dmin is D1, then the remaining distance Drest is D2 and D3. Calculate the average distance Dave of the remaining distance Drest, that is Then the confidence level T is

[0085] For example, (X1, Y1, Z1) = (0, 0, 1), (X2, Y2, Z2) = (0, 1, 0), (X3, Y3, Z3) = (1, 0, 0), (X, Y, Z) = (0.8, 0.2, 0.1), then D1 = = (0-0.8) 2 +(0-0.2) 2 +(1-0.1) 2 =0.89, D2==(0-0.8) 2 +(1-0.2) 2 +(0-0.1) 2 =1.49, D3==(1-0.8) 2 +(0-0.2) 2 +(0-0.1) 2 =1.49, then the minimum distance Dmin=0.89, the average distance Dave of the remaining distance Drest=1.49, then the confidence T=0.40. The smaller the distance value, the closer the predicted solution is to the target solution, that is, the predicted solution belongs to the category corresponding to the target solution D1. With the help of the confidence ranking table and the preset threshold range, the actual solution to the abnormal event is corrected, which can be manual correction or, on the premise that the confidence meets the preset threshold range, the corresponding target solution is automatically selected to replace the actual solution.

[0086] Example 2

[0087] This embodiment provides a solution path prediction system. Figure 3 A schematic diagram of a solution prediction system provided in this embodiment is shown in FIG. Figure 3 As shown, the solution prediction system includes: a verification data set acquisition module 1, which is used to obtain a verification data set, the verification data set includes multiple events, N standard solutions and actual solutions corresponding to each event; the actual solution is one of the N standard solutions; a predicted solution training module 2, which is used to input the verification data set into the initial solution prediction model for training to obtain the predicted solution corresponding to each event; the initial solution prediction model takes the event as input and the predicted solution corresponding to the event as output; a first screening module 3, which is used to screen abnormal events whose predicted solution is inconsistent with the actual solution; a confidence calculation module 4, which is used to calculate the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event; a second screening module 5, which is used to screen several abnormal events whose confidence meets the preset threshold range, and to correct the actual solutions of several abnormal events; a model training module 6, which is used to train the initial solution prediction model based on the corrected verification data set to obtain an optimized solution prediction model.

[0088] The confidence calculation module 4 includes a distance value calculation unit 41, which is used to respectively calculate the distance values ​​between the predicted solution vector corresponding to the predicted solution and the N standard solution vectors corresponding to the N standard solutions; and calculate the confidence based on the distance value.

[0089] The distance value includes the minimum distance Dmin and N-1 remaining distances Drest excluding the minimum distance Dmin; the distance value calculation unit 41 is specifically used to calculate the average distance Dave of the N-1 remaining distances Drest; and according to the formula Calculate the confidence level, where T is the confidence level.

[0090] The preset threshold range includes a first preset threshold range and a second preset threshold range, and the upper limit value of the second preset threshold range is greater than the upper limit value of the first preset threshold range; the second screening module 5 is specifically used to screen a number of abnormal events whose confidence meets the first preset threshold range, and to correct the actual solutions of the several abnormal events; the model training module 6 is specifically used to train the initial solution prediction model based on the corrected verification data set to obtain an optimized solution prediction model; the first screening module 3 is called again to screen abnormal events whose predicted solutions are inconsistent with the actual solutions until the confidence meets the second preset threshold range.

[0091] The solution prediction system of this embodiment includes a verification data set acquisition module, a predicted solution training module, a first screening module, a confidence calculation module, a second screening module and a model training module. Through the mutual cooperation of various modules, the data accuracy of the verification data set in the solution prediction system is improved, the solution prediction model is optimized, the accuracy of the predicted solution is improved, the workload of manual correction of the verification data set is reduced, and the work efficiency is improved.

[0092] Example 3

[0093] This embodiment provides an electronic device, Figure 4 This is a schematic diagram of the structure of an electronic device shown in this embodiment. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the training method of the solution path prediction model in the above-mentioned embodiment 1 is implemented. Figure 4 The electronic device 50 shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention. Figure 4 As shown, the electronic device 50 may be in the form of a general-purpose computing device, for example, it may be a server device. The components of the electronic device 50 may include, but are not limited to: at least one processor 51, at least one memory 52, and a bus 53 connecting different system components (including the memory 52 and the processor 51).

[0094] The bus 53 includes a data bus, an address bus, and a control bus.

[0095] The memory 52 may include a volatile memory, such as a random access memory (RAM) 521 and / or a cache memory 522 , and may further include a read-only memory (ROM) 523 .

[0096] The memory 52 may also include a program tool 525 (or utility) having a set (at least one) of program modules 524, such program modules 524 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0097] The processor 51 executes various functional applications and data processing by running the computer programs stored in the memory 52, such as the training method of the solution path prediction model in the above-mentioned embodiment 1.

[0098] The electronic device 50 may also communicate with one or more external devices 54. Such communication may be performed via an input / output (I / O) interface 55. Furthermore, the model generating electronic device 50 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 56. Figure 4 As shown, the network adapter 56 communicates with other modules of the electronic device 50 via the bus 53. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 50, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0099] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0100] Example 4

[0101] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the training method of the solution path prediction model in the above-mentioned embodiment 1 is implemented.

[0102] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0103] In a possible implementation, the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to enable the terminal device to execute the steps in the training method for implementing the solution path prediction model in the above-mentioned embodiment 1.

[0104] Among them, the program code for executing the present invention can be written in any combination of one or more programming languages, and the program code can be executed completely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or completely on the remote device.

[0105] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that this is only for illustration and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A training method for a solution path prediction model, characterized in that: The training method comprises: Acquire a verification data set, the verification data set including a plurality of events, N standard solutions and an actual solution corresponding to each of the events; the actual solution is one of the N standard solutions; Inputting the verification data set into the initial solution path prediction model for training to obtain the predicted solution paths corresponding to each of the events; the initial solution path prediction model takes the events as input and takes the predicted solution paths corresponding to the events as output; Screen out abnormal events where the predicted solution path does not match the actual solution path; Calculating the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event; Screening a number of abnormal events whose confidences meet a preset threshold range, and revising actual solutions to the number of abnormal events; Training the initial solution pathway prediction model based on the revised validation data set to obtain an optimized solution pathway prediction model; Wherein, the event includes at least one of a computer failure, a network failure or a telephone failure; The step of calculating the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event specifically includes: Respectively calculating the distance values ​​between the predicted solution vector corresponding to the predicted solution and the N standard solution vectors corresponding to the N standard solution; calculating the confidence based on the distance values; The distance value includes a minimum distance Dmin and N-1 remaining distances Drest excluding the minimum distance Dmin; Calculate the average distance Dave of the N-1 remaining distances Drest; The step of calculating the confidence level based on the distance value specifically includes: According to the formula The confidence level is calculated, where T is the confidence level.

2. The training method according to claim 1, characterized in that: The preset threshold range includes a first preset threshold range and a second preset threshold range; an upper limit value of the second preset threshold range is greater than an upper limit value of the first preset threshold range; The step of screening a number of abnormal events whose confidences meet a preset threshold range and correcting actual solutions to the number of abnormal events specifically includes: Screening a number of abnormal events whose confidences meet the first preset threshold range, and revising actual solutions to the number of abnormal events; Training the initial solution pathway prediction model based on the revised validation data set to obtain an optimized solution pathway prediction model; Return to the step of screening abnormal events whose predicted solution paths do not match the actual solution paths, until the confidence level meets the second preset threshold range.

3. A solution path prediction system, characterized in that: The solution prediction system includes: a verification data set acquisition module, which is used to acquire a verification data set, and the verification data set includes multiple An event, N standard solutions and an actual solution corresponding to each of the events; the actual solution being one of the N standard solutions; A prediction solution path training module, used for inputting the verification data set into an initial solution path prediction model for training to obtain a prediction solution path corresponding to each of the events; the initial solution path prediction model takes the event as input and takes the prediction solution path corresponding to the event as output; The first screening module is used to screen abnormal events where the predicted solution path does not match the actual solution path; A confidence calculation module, used to calculate the confidence of the abnormal event based on the N standard solutions and the predicted solution of the abnormal event; A second screening module is used to screen a number of abnormal events whose confidences meet a preset threshold range, and to modify actual solutions to the number of abnormal events; A model training module, used to train the initial solution pathway prediction model based on the revised validation data set to obtain an optimized solution pathway prediction model; Wherein, the event includes at least one of a computer failure, a network failure or a telephone failure; The confidence calculation module comprises: a distance value calculation unit, used to respectively calculate the distance values ​​between the predicted solution path vector corresponding to the predicted solution path and the N standard solution path vectors corresponding to the N standard solution paths; and calculate the confidence based on the distance values; The distance value includes a minimum distance Dmin and N-1 remaining distances Drest excluding the minimum distance Dmin; The distance value calculation unit is specifically used to calculate the average distance Dave of the N-1 remaining distances Drest; and according to the formula The confidence level is calculated, where T is the confidence level.

4. The solution path prediction system according to claim 3, characterized in that: The preset threshold range includes a first preset threshold range and a second preset threshold range; an upper limit value of the second preset threshold range is greater than an upper limit value of the first preset threshold range; The second screening module is specifically used to screen a number of abnormal events whose confidences meet the first preset threshold range, and to modify actual solutions to the number of abnormal events; The model training module is specifically used to train the initial solution path prediction model based on the revised verification data set to obtain an optimized solution path prediction model; The first screening module is called again to screen abnormal events in which the predicted solution path does not match the actual solution path, until the confidence level meets the second preset threshold range.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the training method of the solution path prediction model as claimed in claim 1 or 2 is implemented.

6. A computer readable medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the training method of the solution path prediction model according to claim 1 or 2 is implemented.

Images