A pre-orchestration process processing method based on AI intelligent algorithms

Abstract

The present invention provides a pre-orchestration process processing method based on an AI intelligent algorithm, belonging to the technical field of Internet TV, specifically including: obtaining historical operation behavior logs within a first time threshold, and performing feature extraction based on the historical operation behavior logs to obtain historical operation features; constructing a data set based on the historical operation features, training a prediction model based on unsupervised learning, and obtaining a trained prediction model; real-time collecting log features of operation behavior logs, sending the log features into the prediction model to obtain a final audit result, and outputting corresponding processing suggestions based on the final audit result, thereby further improving the efficiency and accuracy of pre-orchestration process processing.

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet TV, and specifically relates to a pre-orchestration process processing method based on an AI intelligent algorithm. Background Art

[0002] In the IPTV service field, when operators upload media assets on a daily basis, they need to go through operation processes such as media asset production and processing, media asset editing, media asset review, media asset orchestration, media asset release, media asset packaging, and media asset online. For each operation, they need to wait for success before performing the next operation. The entire process is a completely linear process, with very low efficiency, requiring a large number of operators. At the same time, a series of problems such as many review exceptions, release failures, and operation failures will occur during the operation, and professional technical personnel are needed to handle and support these problems.

[0003] In order to implement the review of the process during the orchestration in IPTV, in the authorized invention patent authorization announcement number CN108551592B, "An EPG self-orchestration method and system based on IPTV", by specifying components for different block jumps according to the operation content of the EPG for block display, and updating the review information and program information; previewing the display effect of the EPG, if modification is required, after secondary editing of the blocks in the EPG that need to be modified, after the editing is completed, publishing the EPG, the method of manual review also has the problem of low efficiency. In addition, an intelligent customer service system can also be used, but it only summarizes and precipitates various problems that have been processed in the past, and there is no intelligent solution suggestion for problem handling. Moreover, usually, installing such a system requires high system procurement costs, as well as hardware procurement costs and deployment and maintenance, and it cannot automatically handle related problems.

[0004] Based on the above technical problems, it is necessary to design a pre-orchestration process processing method based on an AI intelligent algorithm. Summary of the Invention

[0005] The purpose of the present invention is to provide a pre-orchestration process processing method based on an AI intelligent algorithm.

[0006] In order to solve the above technical problems, the first aspect of the present invention provides a pre-orchestration process processing method based on an AI intelligent algorithm, specifically including:

[0007] S11 Obtain historical operation behavior logs within a first time threshold, and perform feature extraction based on the historical operation behavior logs to obtain historical operation features;

[0008] S12 Construct a data set based on the historical operation features, train a prediction model based on unsupervised learning, and obtain a trained prediction model;

[0009] S13 Collect the log features of the operation behavior in real time, send the log features into the prediction model to obtain the final audit result, and output the corresponding processing suggestions based on the final audit result.

[0010] Through the construction of the prediction model, it is possible to analyze the media assets of the operators and automatically pre-arrange and generate a process step. The media assets will be automatically processed and transferred according to this process step. At the same time, when there are operation problems, the model can prompt the possible problems and corresponding solutions for the operators. The operators can operate and solve the problems according to the instructions, which greatly improves the efficiency of the operators and reduces the labor costs of the operators and technical support personnel.

[0011] A further technical solution is that the historical operation behavior logs at least include media asset production logs, media asset editing logs, media asset audit logs, media asset distribution logs, media asset process error logs, and logs with successful execution results after errors occur.

[0012] A further technical solution is that the historical operation features further include time identification features, media asset unique identification features, distribution node identification features, media asset type features, and audit results.

[0013] A further technical solution is to split the data set, with 80% of the data set as the training set, 10% of the data set as the test set, and 10% of the data set as the validation set.

[0014] A further technical solution is to construct an updated data set by combining the log features of the operation behavior logs collected in real time with the historical operation features, and continue to train the prediction model based on the updated data set to obtain a new prediction model, and use the new prediction model after the second time threshold to update the existing prediction model to obtain an updated prediction model.

[0015] Through the real-time update of the prediction model, the prediction model can perform continuous learning and training, thereby further enhancing the richness, accuracy, and usability of the intelligent pre-arrangement and recommended solutions it provides.

[0016] A further technical solution is that the final audit result is constructed at least based on the audit results of multiple prediction models within the third time threshold, and the third time threshold is greater than the second time threshold.

[0017] A further technical solution is that the specific steps for constructing the final audit based on the audit results of multiple prediction models within the third time threshold are as follows:

[0018] S21 Send the log features into multiple prediction models within the third time threshold to obtain multiple audit results;

[0019] S22 determines the same prediction model that is the same as the review result of the latest prediction model based on the review result of the latest prediction model, and obtains the accuracy score of the review result based on the accuracy score of the same prediction model and the accuracy score of the review result of the latest prediction model, and determines whether the accuracy score of the review result is greater than the first confidence threshold. If so, it proceeds to step S23; if not, it returns to step S21 to continue constructing the review result.

[0020] S23 takes the review result as the final review result.

[0021] By setting the first confidence threshold, it is possible to further ensure the accuracy of the final review result and prevent the technical problems that the final review result is easily interfered and the result is not accurate enough caused by the review result of a single prediction model.

[0022] A further technical solution is that the calculation formula for the accuracy score of the review result is:

[0023]

[0024] where f i is the accuracy score of the i-th same prediction model that is the same as the review result of the latest prediction model, t i is the time in days of the i-th same prediction model that is the same as the review result of the latest prediction model from the latest prediction model, σ is the weight of the accuracy score of the latest prediction model, f is the accuracy score of the latest prediction model, N is the total number of all same prediction models that are the same as the review result of the latest prediction model, and K1, K2 are constants.

[0025] On the other hand, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed on a computer, the computer is made to execute the above-mentioned pre-orchestration process processing method based on the AI intelligent algorithm.

[0026] On the other hand, an embodiment of the present application provides a computer program product, characterized in that the computer program product stores instructions, and when the instructions are executed by a computer, the computer is made to implement the above-mentioned pre-orchestration process processing method based on the AI intelligent algorithm.

[0027] Other features and advantages will be described in the following description, and some will be obvious from the description, or understood by implementing the present invention. The objectives and other advantages of the present invention are achieved and obtained by the structures specifically pointed out in the description and the drawings.

[0028] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following provides preferred embodiments and detailed descriptions in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] By referring to the accompanying drawings and describing its exemplary embodiments in detail, the above and other features and advantages of the present invention will become more obvious.

[0030] Figure 1 is a flowchart of a pre-orchestration process processing method based on an AI intelligent algorithm according to Embodiment 1;

[0031] Figure 2 is a training sample diagram of media asset review logs according to Example 1;

[0032] Figure 3 is an example diagram of a prompt and suggestion solution for a pre-orchestration process processing method based on an AI intelligent algorithm in Embodiment 1.

[0033] Figure 4 is a diagram of the orchestration result of a pre-orchestration process processing method based on an AI intelligent algorithm in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that the present invention will be thorough and complete, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art. The same reference numerals in the drawings denote the same or similar structures, and thus their detailed descriptions will be omitted.

[0035] The terms "a", "an", "the", and "said" are used to indicate the presence of one or more elements / components / etc.; the terms "comprising" and "having" are used to mean an open inclusion and mean that there may be additional elements / components / etc. in addition to the listed elements / components / etc.

[0036] There are the following two problems in the prior art:

[0037] 1) Generally, the process processing system defines the processing process in advance and then executes it step by step. There are many different situations during the entire process of daily film uploading in operation, and different operations are required. Moreover, the initiated process links are not fully confirmed. Therefore, the general process method cannot meet the requirements of improving the efficiency of daily film uploading in operation and the flexibility required for operations.

[0038] 2) For general daily operation problems, the common approach is to install an intelligent customer service system, but usually the cost is relatively high. Additionally, operation problems usually involve media asset editing, review issues, operation negligence issues, related hardware networks for media asset aggregation and distribution, disk storage issues, as well as data synchronization and flow issues among multiple integrated systems. There is currently no customer service system in the market for this complex scenario. At the same time, the customer service system only provides some handling suggestions and cannot automatically handle related problems.

[0039] Embodiment 1

[0040] To solve the above problems, according to one aspect of the present invention, as Figure 1 shown, a pre-orchestration process processing method based on AI intelligent algorithms is provided, specifically including:

[0041] S11 Obtain historical operation behavior logs within a first time threshold, and perform feature extraction based on the historical operation behavior logs to obtain historical operation features;

[0042] It should be noted additionally that,

[0043] 1. Collect media asset production logs for one year, and the specific time can be adjusted.

[0044] Specifically, capture: logs of log type ProductionLog, and extract field values of result media asset fields, desc result description values, desc2 detailed result processing description values, code unique media asset identifier, name media asset name, updatetime operation time, etc. from the logs as the main media asset processing features. At the same time, Chinese word segmentation will be performed on the desc2 detailed result processing description values to extract media asset production features and used to calculate the correlation between relevant features.

[0045] 2. Collect media asset editing logs for one year, and the specific time can be adjusted.

[0046] Specifically, capture: logs of log type EditLog, and extract field values of result media asset fields, desc result description values, desc2 detailed result processing description values, code unique media asset identifier, name media asset name, actor, updatetime operation time, etc. from the logs as the main log features of media asset editing. At the same time, Chinese word segmentation will be performed on the desc2 detailed result processing description values to extract editing result features and used to calculate the correlation between relevant features. At the same time, for changed media asset fields, such as name, actor, etc., the changed values will be extracted by separating with delimiter ->>. The change of field values corresponds to the change of media asset field features.

[0047] 3. Collect media asset review logs for one year, and the specific time is adjustable.

[0048] Specifically, capture logs with the log type of AuditLog, and extract the following field values from the logs as the main log features for media asset editing: the media asset field value of the result, the result description value of desc, the detailed result processing description value of desc2, the unique identifier of the media asset identity code, the media asset name, the operation time of updatetime, etc. At the same time, Chinese word segmentation will be performed on the detailed result processing description value of desc2 to extract the review result features and use them to calculate the correlation between relevant features.

[0049] 4. Collect media asset distribution logs for one year, and the specific time is adjustable.

[0050] Specifically, capture logs with the log type of AuditLog, and extract the following field values from the logs as the main log features for media asset editing: the media asset field value of the result, the result description value of desc, the detailed result processing description value of desc2, the unique identifier of the media asset identity code, the media asset name, the operation time of updatetime, etc. At the same time, Chinese word segmentation will be performed on the detailed result processing description value of desc2 to extract the review result features and use them to calculate the correlation between relevant features.

[0051] 5. Collect media asset process error logs and logs with successful execution results after errors occur, and the specific time is adjustable.

[0052] Specifically, capture logs with the log type of DistributionLog, and extract the following field values from the logs as the main log features for media asset editing: the media asset field value of the result, the result description value of desc, the detailed result processing description value of desc2, the unique identifier of the media asset identity code, the media asset name, the distribution domain field value of outPassageName, the operation time of updatetime, etc. At the same time, Chinese word segmentation will be performed on the detailed result processing description value of desc2 to extract the distribution result features and use them to calculate the correlation between relevant features.

[0053] As Figure 2 shown, it is a training sample diagram of media asset review logs.

[0054] S12 Build a data set based on the historical operation features, train a prediction model based on unsupervised learning, and obtain the trained prediction model;

[0055] It should be noted that for the prediction model based on unsupervised learning, specifically using bottom-up unsupervised learning, training starts from the bottom layer and progresses layer by layer to the top layer. TensorFlow will first learn the parameters of the first layer based on the training set sample data. This layer can be regarded as a hidden layer of a three-layer neural network that minimizes the difference between the output and the input. Due to the limitations of the model's capabilities and sparsity constraints, the resulting model can learn the structure of the data itself, thereby obtaining features with stronger representational ability than the input. Through continuous iterative training, the loss function becomes smaller and smaller until the expected threshold is reached, and then one layer is obtained. After learning the (n - 1)-th layer, the output of the (n - 1)-th layer is used as the input of the n-th layer to train the n-th layer, and the parameters of each layer are obtained respectively. At the same time, the initial model is obtained: AIdingdangModel.temp4) Input the validation dataset ValidationSet through top-down supervised learning, and based on the parameters obtained in the first step, finely adjust the parameters of the multi-layer model and the initial model AIdingdangModel.temp, and then perform supervised training. During the training process, the neural model is updated based on the dynamic update algorithm. The dynamic update algorithm uses conventional weight updates, and the calculation formula is:

[0056]

[0057] So that the weight update can change with the number of iterations. t represents the current number of iterations, and tmax is the maximum number of iterations.

[0058] It should be noted that the entire algorithm uses the gradient descent solution method for the derivative calculation process. Since the first step of deep learning is not random initialization but obtained by learning the structure of the input data, this initial value is closer to the global optimum, thus being able to achieve better results.

[0059] Specifically, take an example of media asset review:

[0060] Part of the imported log after being cleared is:

[0061] Example 1:

[0062] Data-1 = {"result":"1","desc":"sucess","desc2":"Review successful","code":"HBGD6961632939043430405692305035","name":"Monkey King: Hero is Back","updatetime":"2022-10-23:15:55:11.226"} import-test-data(Data-1)

[0063] The features obtained are: media feature: The Return of the Great Sage, identity feature: HBGD6961632939043430405692305035

[0064] Time characteristics: 2022-10-23:15:55:11.226

[0065] Audit result characteristics: Audit successful

[0066] S13 collects log features of the operation behavior log in real time, and sends the log features to the prediction model to obtain the final audit result, and outputs corresponding processing suggestions based on the final audit result.

[0067] It should be noted that the prediction model obtained through unsupervised learning is: AIdingdangModel

[0068] like Figure 3 As shown in the figure, at this time, the real-time collection of operational behavior logs is transmitted to the AIdingdangModel, which can intelligently provide corresponding processing suggestions for the output.

[0069] By constructing a predictive model, it is possible to analyze the media resources of operators and automatically pre-arrange a process step. The media resources will be automatically processed and circulated according to this process step. At the same time, when operational problems arise, the model can prompt operators of possible problems and corresponding solutions. Operators can operate and solve problems according to the instructions, which greatly improves the efficiency of operators and reduces the manpower costs of operators and technical support personnel.

[0070] It should be noted that the historical operation behavior log at least includes a media asset production log, a media asset editing log, a media asset review log, a media asset distribution log, a media asset process error log, and a log of successful execution results after an error occurs.

[0071] In another possible embodiment, the historical operation characteristics further include a time identification feature, a media asset unique identification feature, a distribution node identification feature, a media asset type feature, and an audit result.

[0072] In another possible embodiment, the data set is split, with 80% of the data set used as a training set, 10% of the data set used as a test set, and 10% of the data set used as a validation set.

[0073] It should be noted that the five types of log data are cleaned, invalid data is removed, and sample logs that need deep learning training are retained. The five types of data are split into data sets. For each type of data, 80% of the data set is used as a training set, 10% of the data set is used as a test set, and 10% of the data set is used as a validation set.

[0074] Data cleaning set:

[0075] DataCleanList = f(clean: ProductionLog) + f(clean: EditLog) + f(clean: AuditLog) + f(clean: DistributionLog) + f(clean: ErrorLog)

[0076] Training set: TrainingSet = f(extract: DataCleanList * 80%);

[0077] Test set: TestSet = f(extract: DataCleanList * 10%);

[0078] Validation set: ValidationSet = f(extract: DataCleanList * 10%);

[0079] In another possible embodiment, the log features of the operation behavior logs obtained by real-time collection are combined with the historical operation features to construct an updated data set, and the prediction model is continuously trained based on the updated data set to obtain a new prediction model, and the existing prediction model is updated using the new prediction model after the second time threshold to obtain an updated prediction model.

[0080] It should be noted additionally that 1) an open-source framework TensorFlow is used to build a neural network.

[0081] Specifically, the import tensorflow module is imported. The training set TrainingSet and test set TestSet of the specified network are input. Then the network structure is built layer by layer.

[0082] Make a copy of the model and name it AIdingdangModel.snapshot. Collect the real-time collected logs and the results of log execution and import them into AIdingdangModel.snapshot for continuous learning and training.

[0083] At the same time, AIdingdangModel can automatically generate suggestions for pre-orchestration process steps based on the batch media asset data provided by the operation personnel. The operation personnel can directly execute it with one click or make some detailed adjustments before execution. The logs of the pre-orchestration process will also be collected and imported into AIdingdangModel.snapshot for continuous training.

[0084] Cover the AIdingdangModel model with the continuously learning and training AIdingdangModel.snapshot every week. In this way, iterate and train the AIdingdangModel model on a weekly basis to improve the richness, accuracy, and usability of the intelligent pre-orchestration and suggested solutions it provides.

[0085] By continuously updating the prediction model in real time, the prediction model can continuously learn and train, thereby further improving the richness, accuracy, and usability of the intelligent pre-orchestration and suggested solutions it provides.

[0086] In another possible embodiment, the final review result is constructed at least based on the review results of multiple prediction models within a third time threshold, and the third time threshold is greater than the second time threshold.

[0087] It should be noted that the continuously learning and training model will train the latest 5 versions of AIdingdangModel simultaneously, give the accuracy score for making decisions based on the logs, and use the comprehensive weights to learn the accuracy of the prediction results to provide them to the currently training model for use. Example:

[0088] F(score:logset) = f(AIdingdangModel - 2022 - 10 - 09:logset) * σ1 +

[0089] f(AIdingdangModel - 2022 - 10 - 16:logset) * σ2 +

[0090] f(AIdingdangModel - 2022 - 10 - 23:logset) * σ3 +

[0091] f(AIdingdangModel - 2022 - 10 - 30:logset) * σ4 +

[0092] f(AIdingdangModel - 2022 - 11 - 06:logset) * σ5

[0093] Among them, for the weight assignment, the model version trained on the most recent date has a higher weight, which decreases sequentially according to the historical version time.

[0094] σ5 = 30%

[0095] σ4 = 25%

[0096] σ3 = 20%

[0097] σ2 = 15%

[0098] σ1 = 10%

[0099] Denoted as the formula:

[0100] AIdingdangModel.snapshot = f(AIdingdangModel.snapshot: f(score: logset))

[0101] Where logset is the set of imported behavior logs and result logs.

[0102] f(score: logset) imports logset to calculate the corresponding set of scores for behavior logs and results.

[0103] The corresponding set of scores for behavior logs and results of f(AIdingdangModel.snapshot: f(score: logset)) is imported into the training model of continuous learning for training to obtain a new model for learning iteration.

[0104] If you want to obtain extreme accuracy, you can perform comprehensive weighted training decision scoring on all historical versions of the model. However, more computing power needs to be invested.

[0105] The formula is denoted as:

[0106] f(score: logset) = ∑(i = 1, n = AIdingdangModel... version n) f(score: AIdingdangModel... version i) * σi / n

[0107] Import the AIdingdangModel models of N versions into the set of logs of logset behavior and results for calculation, and then obtain the overall average score.

[0108] In another possible embodiment, the specific steps for constructing the final review based on the review results of multiple prediction models within the third time threshold are as follows:

[0109] S21 Send the log features into multiple prediction models within the third time threshold to obtain multiple review results;

[0110] S22 Based on the review result of the latest prediction model, determine the same prediction models that are the same as the review result of the latest prediction model, and based on the accuracy score of the same prediction model and the accuracy score of the review result of the latest prediction model, obtain the accuracy score of the review result, and determine whether the accuracy score of the review result is greater than the first confidence threshold. If so, enter step S23; if not, return to step S21 to continue constructing the review result;

[0111] S23 uses the review result as the final review result.

[0112] It should be noted that, as Figure 4 shown, it is an example diagram of the pre - arranged automatic processing flow generated by the AIdingdangModel intelligent generation.

[0113] By setting the first confidence threshold, it is possible to further ensure the accuracy of the final review result, and prevent the technical problems that the final review result is easily interfered and the result is inaccurate caused by the review result of a single prediction model.

[0114] In another possible embodiment, the calculation formula of the accuracy score of the review result is:

[0115]

[0116] where f i is the accuracy score of the i - th identical prediction model with the same review result as the latest prediction model, t i is the time in days from the i - th identical prediction model with the same review result as the latest prediction model to the latest prediction model, σ is the weight of the accuracy score of the latest prediction model, f is the accuracy score of the latest prediction model, N is the total number of all identical prediction models with the same review result as the latest prediction model, and K1, K2 are constants.

[0117] Embodiment 2

[0118] An embodiment of the present application provides a computer - readable storage medium, on which a computer program is stored. When the computer program is executed on a computer, the computer is made to execute the above - mentioned pre - arranged process processing method based on an AI intelligent algorithm.

[0119] Embodiment 3

[0120] An embodiment of the present application provides a computer program product, characterized in that the computer program product stores instructions, and when the instructions are executed by a computer, the computer is made to implement the above - mentioned pre - arranged process processing method based on an AI intelligent algorithm.

[0121] In several embodiments provided by the present application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of systems, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and a module, a program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0122] In addition, in each embodiment of the present invention, the functional modules can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part.

[0123] If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0124] Based on the above inspiration from the ideal embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A pre-orchestration process processing method based on AI intelligent algorithms, characterized in that, Specifically, it includes: S11 Obtain historical operation behavior logs within the first time threshold, and perform feature extraction based on the historical operation behavior logs to obtain historical operation features; S12 Construct a data set based on the historical operation features, train a prediction model based on unsupervised learning, and obtain a trained prediction model; S13 Real-time collect the log features of operation behavior logs, and send the log features into the prediction model to obtain the final review result, and output corresponding processing suggestions based on the final review result; Construct an updated data set together with the log features of the operation behavior logs obtained by the real-time collection and the historical operation features, and continue to train the prediction model based on the updated data set to obtain a new prediction model, and use the new prediction model after the second time threshold to update the existing prediction model to obtain an updated prediction model; S21 Send the log features into multiple prediction models within the third time threshold to obtain multiple review results; S22 Based on the review result of the latest prediction model, determine the same prediction models with the same review result as the review result of the latest prediction model, and based on the accuracy score of the same prediction model and the accuracy score of the review result of the latest prediction model, obtain the accuracy score of the review result, and determine whether the accuracy score of the review result is greater than the first confidence threshold. If so, go to step S23; if not, return to step S21 to continue constructing the review result; S23 Take the review result as the final review result.

2. The preorchestrated process processing method according to claim 1, characterized in that, The historical operation behavior logs at least include media asset production logs, media asset editing logs, media asset review logs, media asset distribution logs, media asset process error logs, and logs with successful execution results after errors occur.

3. The pre-arranged process processing method according to claim 1, characterized in that, The historical operation features also include time identification features, media asset unique identification features, distribution node identification features, media asset type features, and review results.

4. The pre-arranged process processing method according to claim 1, wherein Split the data set, with 80% of the data set as the training set, 10% of the data set as the test set, and 10% of the data set as the validation set.

5. The pre-arranged process processing method according to claim 1, wherein The final review result is constructed at least based on the review results of multiple prediction models within the third time threshold, and the third time threshold is greater than the second time threshold.

6. The pre-orchestrated process processing method according to claim 1, wherein The calculation formula for the accuracy score of the review result is: where f i is the accuracy score of the i-th identical prediction model with the same review result as the latest prediction model, t i is the time in days of the i-th identical prediction model with the same review result as the latest prediction model from the latest prediction model, σ is the weight of the accuracy score of the latest prediction model, f is the accuracy score of the latest prediction model, N is the total number of all identical prediction models with the same review result as the latest prediction model, and K1, K2 are constants.

7. A computer-readable storage medium, on which a computer program is stored. When the computer program is executed in a computer, the computer is made to execute a pre-orchestration process processing method based on an AI intelligent algorithm according to any one of claims 1-6.

8. A computer program product, characterized in that, The computer program product stores instructions, and when the instructions are executed by a computer, the computer is made to implement a pre-orchestration process processing method based on an AI intelligent algorithm according to any one of claims 1-6.

Images