Training method and apparatus for task execution time-series prediction model

Abstract

This application provides a method and apparatus for training a task execution time series prediction model. The method includes: collecting historical task execution information of a target organization; constructing a sample time series based on the historical task execution information; performing sequence decomposition on the sample time series to obtain seasonal and trend terms; sequentially inputting the seasonal and trend terms and the sample time series into an initial prediction model, outputting first, second, and third prediction information respectively; constructing seasonal, trend, and overall losses based on the first, second, and third prediction information and the sample time series; constructing a hybrid loss function based on the seasonal, trend, and overall losses; and training the initial prediction model using the hybrid loss function to obtain a target prediction model. The obtained target prediction model, without affecting computational efficiency, performs high-precision prediction of the target organization's task execution information, obtaining a balanced task execution plan.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for training a task execution time-series prediction model. This application also relates to a task execution time-series prediction model training apparatus, a computing device, and a computer-readable storage medium. Background Technology

[0002] As various operations within existing organizations continue to expand, especially with the increasing number of tasks in the public sector, the distribution of tasks within the same cycle is crucial for task handlers. For example, within an annual task execution cycle, if some quarters have a high concentration of tasks while others have a low concentration, it will lead to a decline in the quality of tasks completed in the former and low efficiency in the latter. Therefore, there is an urgent need for a method to achieve balanced task execution planning. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method for training a task execution timing prediction model to address the technical deficiencies in the prior art. Embodiments of this application also provide a task execution timing prediction model training apparatus, a computing device, and a computer-readable storage medium.

[0004] According to a first aspect of the embodiments of this application, a method for training a task execution timing prediction model is provided, comprising:

[0005] Collect historical task execution information of the target organization, and construct a sample time series based on the historical task execution information;

[0006] The sample time series is decomposed to obtain the seasonal and trend terms associated with the sample time series;

[0007] The seasonal term is input into the initial prediction model, and the first prediction information is output. The trend term is input into the initial prediction model, and the second prediction information is output. The sample time series is input into the initial prediction model, and the third prediction information is obtained.

[0008] Based on the first prediction information, the second prediction information, the third prediction information, and the sample time series, seasonal loss, trend loss, and overall loss are constructed.

[0009] Based on the seasonal loss, the trend loss, and the overall loss, a hybrid loss function is constructed, and the initial prediction model is trained using the hybrid loss function to obtain the target prediction model.

[0010] Optionally, the initial prediction model includes an attention mechanism layer, wherein the influence factor associated with the attention mechanism layer is the task type of the task data contained in the historical task execution information.

[0011] Optionally, the step of inputting the seasonal term into the initial prediction model and outputting the first prediction information includes:

[0012] The seasonal features are extracted by using the convolutional layer of the initial prediction model to obtain the target seasonal features;

[0013] The target seasonal features are reduced in dimensionality by using the pooling layer of the initial prediction model to obtain the dimensionality-reduced seasonal features.

[0014] The attention mechanism layer processes the dimensionality-reduced seasonal features, and the processing result is input into the fully connected layer of the initial prediction model for further processing to obtain the first prediction information. During the data processing of the initial prediction model, the dropout layer of the initial prediction model randomly discards neurons of the initial prediction model.

[0015] Optionally, constructing a sample time series based on the historical task execution information includes:

[0016] Query the missing values ​​in the historical task execution information and fill in the missing values ​​using a preset interpolation algorithm;

[0017] Under the assumption of normal distribution, based on the degree of variation in statistical data, we sequentially determine whether the task data contained in the historical task execution information are outliers.

[0018] If so, remove the outlier and supplement the outlier according to the interpolation algorithm;

[0019] If not, retain the task data until all task data has been processed to obtain sample task execution information, and construct the sample time series based on the sample task execution information.

[0020] Optionally, constructing seasonal loss, trend loss, and overall loss based on the first prediction information, the second prediction information, the third prediction information, and the sample time series includes:

[0021] The seasonal loss is constructed based on the first prediction information and the sample time series;

[0022] The trend loss is constructed based on the second prediction information and the sample time series;

[0023] The overall loss is constructed based on the third prediction information and the sample time series.

[0024] Optionally, constructing a hybrid loss function based on the seasonal loss, the trend loss, and the overall loss includes:

[0025] Based on a preset first weighting coefficient, the seasonal loss and the trend loss are weighted and summed to obtain the component loss function;

[0026] Based on a preset second weighting coefficient, the component loss function and the overall loss are weighted and summed to obtain a hybrid loss function.

[0027] Optionally, training the initial prediction model using the hybrid loss function to obtain the target prediction model includes:

[0028] Based on the preset min-max method, the first weight coefficient and the second weight coefficient are optimized to obtain the target mixed loss function;

[0029] The initial prediction model is trained based on the target mixture loss function to obtain the target prediction model.

[0030] According to a second aspect of the embodiments of this application, a task execution timing prediction model training apparatus is provided, comprising:

[0031] The acquisition module is configured to acquire historical task execution information of the target organization and construct a sample time series based on the historical task execution information;

[0032] The decomposition module is configured to perform sequence decomposition on the sample time series to obtain the seasonal and trend terms associated with the sample time series.

[0033] The processing module is configured to input the seasonal term into the initial prediction model and output first prediction information, input the trend term into the initial prediction model and output second prediction information, and input the sample time series into the initial prediction model to obtain third prediction information;

[0034] The construction module is configured to construct seasonal loss, trend loss, and overall loss based on the first prediction information, the second prediction information, the third prediction information, and the sample time series.

[0035] The training module is configured to construct a hybrid loss function based on the seasonal loss, the trend loss, and the overall loss, and to train the initial prediction model using the hybrid loss function to obtain the target prediction model.

[0036] According to a third aspect of the embodiments of this application, a computing device is provided, comprising:

[0037] Memory and processor;

[0038] The memory is used to store computer-executable instructions, and the processor executes the computer-executable instructions to implement the steps of the task execution timing prediction model training method.

[0039] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the task execution timing prediction model training method.

[0040] According to a fifth aspect of the present application, a chip is provided that stores a computer program, which, when executed by the chip, implements the steps of the task execution timing prediction model training method.

[0041] The task execution time-series prediction model training method provided in this application collects historical task execution information of the target organization and constructs a sample time series based on the historical task execution information; performs sequence decomposition on the sample time series to obtain seasonal and trend terms associated with the sample time series; inputs the seasonal terms into an initial prediction model to output first prediction information, inputs the trend terms into the initial prediction model to output second prediction information, and inputs the sample time series into the initial prediction model to obtain third prediction information; constructs seasonal loss, trend loss, and overall loss based on the first, second, and third prediction information and the sample time series; constructs a hybrid loss function based on the seasonal, trend, and overall losses, and trains the initial prediction model using the hybrid loss function to obtain a target prediction model. The obtained target prediction model achieves high-precision prediction of the target organization's task execution information without affecting computational efficiency, obtaining a balanced task execution plan. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a task execution timing prediction model training method provided in an embodiment of this application;

[0044] Figure 2 This is a schematic diagram of a training method for a task execution timing prediction model provided in an embodiment of this application;

[0045] Figure 3This is a schematic diagram of the structure of a task execution timing prediction model training device provided in an embodiment of this application;

[0046] Figure 4 This is a structural block diagram of a computing device provided in one embodiment of this application. Detailed Implementation

[0047] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0048] The terminology used in one or more embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the scope of one or more embodiments of this application. The singular forms “a,” “the,” and “the” used in one or more embodiments of this application and in the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more associated listed items.

[0049] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this application, and similarly, second may also be referred to as first.

[0050] This application provides a method for training a task execution timing prediction model. This application also relates to a task execution timing prediction model training apparatus, a computing device, and a computer-readable storage medium, which will be described in detail in the following embodiments.

[0051] Figure 1 The flowchart illustrates a method for training a task execution timing prediction model according to an embodiment of this application, which specifically includes the following steps:

[0052] Step S102: Collect historical task execution information of the target organization, and construct a sample time series based on the historical task execution information;

[0053] Step S104: Perform sequence decomposition on the sample time series to obtain the seasonal and trend terms associated with the sample time series;

[0054] Step S106: Input the seasonal term into the initial prediction model and output the first prediction information; input the trend term into the initial prediction model and output the second prediction information; input the sample time series into the initial prediction model and obtain the third prediction information.

[0055] Step S108: Construct seasonal loss, trend loss, and overall loss based on the first prediction information, the second prediction information, the third prediction information, and the sample time series;

[0056] Step S110: Construct a hybrid loss function based on the seasonal loss, the trend loss, and the overall loss, and train the initial prediction model using the hybrid loss function to obtain the target prediction model.

[0057] The target institution can be a court, law firm, educational institution, etc., and the specific type of institution is determined by the actual use scenario. This embodiment does not limit this. The historical task execution information represents the task allocation of the target institution over the years, including the task type and the timestamp data corresponding to each task. The process of decomposing the sample time series can be carried out by spectral analysis, time series analysis, and Fourier series analysis, etc. The specific decomposition method adopted is determined by the actual use scenario. This embodiment does not limit this. The initial model can be a neural network model with a multilayer perceptron (MLP) architecture, such as the Dlinear model, or a neural network model with a Transformer architecture, such as the FEDformer model, PatchTST model, etc. The specific model type selected is determined by the actual use scenario. This embodiment does not limit this.

[0058] Based on this, historical task execution information containing the task allocation of the target organization over the years is collected. A sample time series is constructed based on the collection results. The sample time series is decomposed into seasonal and trend terms. The seasonal, trend, and sample time series are then input into the initial model. The initial model outputs the first, second, and third prediction information for the input seasonal, trend, and sample time series, respectively.

[0059] Based on the three predictions output by the initial model, a loss function corresponding one-to-one with each of these three predictions is constructed. A hybrid loss function is then built based on these three predictions. The initial prediction model is trained using this hybrid loss function to obtain the target prediction model. For task planning within a preset time period of the target organization, such as the next year, the task data to be processed within this preset time period can be input into the target prediction model. The target prediction model outputs a target time series, which represents the processing time points corresponding to the task data to be processed. The target time series ensures a balanced distribution of tasks to be processed within the preset time period.

[0060] Furthermore, in step S102, the process of constructing a sample time series based on historical task execution information is as follows:

[0061] The missing values ​​in the historical task execution information are queried and filled in using a preset interpolation algorithm. Under the assumption of normal distribution, based on the degree of variation of statistical data, the task data contained in the historical task execution information are sequentially judged to see if they are outliers. If so, the outliers are removed and filled in according to the interpolation algorithm. If not, the task data is retained until all the task data is processed to obtain sample task execution information. The sample time series is then constructed based on the sample task execution information.

[0062] Among these steps, the collected historical task execution information needs to be preprocessed, specifically, the sequence needs to be checked for missing values ​​and outliers.

[0063] Therefore, for missing values, linear or polynomial interpolation algorithms are used to fill in the missing values. Outlier detection, on the other hand, is based on statistical methods. Specifically, under the assumption of normal distribution, the degree of variation in statistical data is used to determine whether the data is acceptable, such as 3-Sigma, which uses three standard deviations for detection. In the 3-Sigma method, a band is formed by extending three standard deviations outwards from the mean. If a sample data point falls within this band, the data is considered normal; if it falls outside the band, the data is considered outlier.

[0064] Secondly, the outlier detection process can also determine the seasonal trend of historical task execution information. For data with a clear seasonal trend, the SARIMA model is used to detect outliers. The SARIMA model mainly consists of three parts: seasonal differencing, non-seasonal differencing, and the ARMA model. Seasonal and non-seasonal differencing smooth the time series data, while the ARMA model describes the time series data corresponding to historical task execution information. Its autocorrelation and moving average properties are used to detect outliers by comparing observed data points with the range predicted by the model. If the result is within the predicted range, the data is retained; otherwise, interpolation methods for missing values ​​are used to replace them.

[0065] Furthermore, it is necessary to detect whether the time nodes in the historical task execution information are evenly spaced. Specifically, this can be determined by observing the corresponding timestamps or calculating the time distance between two adjacent time nodes. Since most neural network models assume that the time nodes of the data are evenly distributed, and for historical task execution information that is consistent over a short period of time, the model can more easily learn the patterns and trends of the corresponding time series, thus ensuring that the model has good generalization and interpretability.

[0066] Furthermore, for training the initial prediction model, historical task execution information can be divided according to a specified time step, batch size, and partition ratio to obtain a training set, a test set, and a validation set. The training set consists of sample time series data. Specifically, the optimal partition ratio is 7:1:2, ensuring that 20% of the data is used for testing, achieving sufficient validation during the model evaluation phase to obtain more reliable metrics, while the 10% validation set is used for optimization within a wider parameter range. The partitioning process is implemented by using a sliding window size as a sample, with the batch size corresponding to the number of samples extracted from each batch.

[0067] After determining the sample time series, step S104 requires further decomposition of the sample time series. The decomposition process employs seasonal trend decomposition, i.e., moving average decomposition, for the sample time series X. t It can be decomposed into S t and T t S t and T t These represent the seasonal and trend terms of the sequence, respectively, with each sequence maintaining the same size as the original sequence.

[0068] To ensure the initial prediction model can correctly utilize the processed data, further data transformation is necessary. Standard normalization is used for scaling, transforming the data to the 0-1 range to avoid the impact of differences in the units of measurement between different variables on model training. Further processing is required for different model frameworks. For example, for models using a patch architecture, the sequences need to be converted to the required format according to the patch size, with each patch serving as a token as input to the model. For models not using patches, each time point is input as a token.

[0069] Furthermore, the initial prediction model includes an attention mechanism layer, wherein the influence factor associated with the attention mechanism layer is the task type of the task data contained in the historical task execution information.

[0070] Furthermore, in step S106, the process of inputting the seasonal term into the initial prediction model and outputting the first prediction information is specifically implemented as follows in this embodiment:

[0071] The initial prediction model extracts seasonal features from the convolutional layer to obtain target seasonal features; it then reduces the dimensionality of these features using the pooling layer to obtain dimensionality-reduced seasonal features; finally, it processes these dimensionality-reduced seasonal features using the attention mechanism layer and inputs the processing result into the fully connected layer of the initial prediction model for further processing to obtain the first prediction information. During the initial prediction model data processing, the initial prediction model's dropout layer randomly discards neurons from the initial prediction model.

[0072] The initial prediction model includes convolutional layers, pooling layers, an attention mechanism layer, fully connected layers, and dropout layers. Regarding the processing of seasonal terms in the initial prediction model, features are first extracted using convolutional layers, then the pooling layers reduce the dimensionality of the extracted features. The attention mechanism layer processes the dimensionality reduction result according to the task type associated with the seasonal terms, ultimately obtaining the first prediction information. It should be noted that the processing of trend terms and sample time series data follows the same procedure. Furthermore, as... Figure 2 The provided training method for a task execution timing prediction model is illustrated in the model diagram. In addition to single-time deployment, convolutional layers and max-pooling layers can be deployed in multiple cyclical layers. Dropout layers are also deployed to randomly drop some neurons, thereby improving the robustness of the neural network model.

[0073] For example, regarding key issues in the entire court case-handling process, there are instances of cases not being accepted and a balanced distribution of accepted cases. Precisely monitoring whether these behaviors occur within the courts is crucial for judicial oversight. In actual case-handling procedures, a large amount of statistical data is typically collected regarding the case filing process. This data includes the number of cases filed for various types, including civil, criminal, and administrative cases, and can even be further subdivided. By collecting case filing information for different courts and different types of cases at fixed time steps, this data is compiled into a multivariate time series dataset, i.e., the sample time series X. t The corresponding seasonal term S is obtained by decomposing it through seasonal trends. t and trend item T t S t T t and X t The information is input into the initial prediction model to obtain the first prediction information, the second prediction information, and the third prediction information.

[0074] Furthermore, in step S108, the process of constructing seasonal loss, trend loss, and overall loss based on the first prediction information, the second prediction information, the third prediction information, and the sample time series is specifically implemented as follows in this embodiment:

[0075] The seasonal loss is constructed based on the first prediction information and the sample time series; the trend loss is constructed based on the second prediction information and the sample time series; and the overall loss is constructed based on the third prediction information and the sample time series.

[0076] Continuing with the previous example, based on the first prediction information and the sample time series, the seasonal loss is determined. S Based on the second prediction information and the sample time series, the trend loss is determined. T Based on the third prediction information and the sample time series, the overall loss is determined. G .

[0077] Furthermore, in step S110, the process of constructing a hybrid loss function based on seasonal loss, trend loss, and overall loss is specifically implemented as follows in this embodiment:

[0078] Based on a preset first weighting coefficient, the seasonal loss and the trend loss are weighted and summed to obtain a component loss function; based on a preset second weighting coefficient, the component loss function and the overall loss are weighted and summed to obtain a mixed loss function.

[0079] Furthermore, in step S110, the process of training the initial prediction model using a hybrid loss function to obtain the target prediction model is specifically implemented as follows in this embodiment:

[0080] The first weight coefficient and the second weight coefficient are optimized according to the preset min-max method to obtain the target mixture loss function; the initial prediction model is trained according to the target mixture loss function to obtain the target prediction model.

[0081] Continuing with the previous example, since the influence of the component loss function related to historical task execution information on the overall loss function, as well as the seasonal loss and trend loss, cannot be determined before the initial prediction model training is complete, the component loss function and the overall loss function are given the same emphasis, and the seasonal loss and trend loss are also given the same emphasis. Therefore, the first weight coefficient for the seasonal loss and trend loss is set to... Similarly, the second weighting coefficients for the component loss function and the overall loss function are set to... During subsequent training, the weight ratio of the two sets of weight coefficients is dynamically adjusted, but it must be ensured that the sum of the two sets of weights is a fixed value of 1.

[0082] Regarding the dynamic adjustment process, two min-max optimization objectives are used to balance seasonal and trend losses, as well as to balance component loss functions and seasonal loss functions. Specifically, regarding the component loss function Loss... C With the overall loss function Loss G The optimization target is calculated using the following formula:

[0083]

[0084] Where θ represents the parameters of the initial prediction model corresponding to deep learning, and i is 1 or 2. The above is achieved through the overall loss function Loss. G and component loss function Loss C The magnitudes of ω1 and ω2 are adjusted to maximize the optimized mixture loss function for the current sample. Regarding the seasonal loss... S Trend Loss T The optimization target is calculated using the following formula:

[0085]

[0086] Similarly, the above seasonal losses S Trend Loss T The magnitude of α and β is used to adjust the two parameters, so that the optimized component loss function Loss CThe maximum value is found in the current sample. Then, the losses calculated from the two optimization objectives are combined to obtain the objective hybrid loss function, as shown in the following formula:

[0087]

[0088] The objective mixed loss function is used as the initial prediction model. The mixed loss needs to be learned last. The model weight parameters θ are learned through the backpropagation process of the model, and the model parameters are adjusted to minimize the maximum mixed loss.

[0089] To solve the above optimization problem, gradient descent is used, specifically mirror gradient descent to update the parameters. In particular, the parameters ω1, ω2, α, and β are optimized based on the calculated loss. The specific calculation method is as follows:

[0090]

[0091] Where pre represents the previous update step, and cur represents the current update step. λ i (i = 1 or 2) is a hyperparameter that characterizes the importance of the balancing weights, and its value ranges from [0, 1]. Through the above calculations, the optimization process becomes: for each optimization step, calculate the corresponding parameter value based on the current loss magnitude, resulting in the following mixed loss function, Loss:

[0092] Loss=ω1Loss G +ω2(αLoss S +βLoss T ).

[0093] The target prediction model obtained through the above methods can accurately predict the patterns of court case filing. For example, during holidays such as Spring Festival and National Day, the number of cases filed will decrease significantly due to staff taking leave, further leading to a significant difference in the number of cases in the current month compared to other months due to the holiday effect. In addition, in order to ensure the number of cases received in December, the number of new cases filed will be controlled, resulting in a significant decrease in the number of cases filed at the end of the year. All of these can be reflected in the prediction results of the target prediction model.

[0094] Furthermore, the prediction results of the target prediction model can also reflect the continuous advancement of the rule of law, the increasing legal awareness and litigation awareness of the public, which leads to a year-on-year increase in the number of cases filed. In addition, economic development is accompanied by an increase in social conflicts. Specifically, with economic development, civil and commercial disputes, intellectual property cases, etc., are gradually increasing, leading to a continuous increase in the number of cases filed.

[0095] Furthermore, the prediction results of the target prediction model can also reflect the trends of increase or decrease in the number of cases due to temperature differences between summer and winter, as well as the concentrated outbreak of certain types of cases in specific years, such as the correlation between real estate disputes and economic cyclical fluctuations.

[0096] In summary, the task execution time series prediction model training method in this embodiment, by integrating the relationship between the overall sequence and the component sequences, achieves a focus on more granular components. Based on the temporal characteristics of the components, it improves the prediction effect of the components, thereby enhancing the overall prediction effect and obtaining a target prediction model with high prediction accuracy. Furthermore, compared with traditional time series prediction methods, by simply embedding additional loss construction methods, the impact on the model's data processing volume is minimal, and it will not significantly affect the model's running efficiency.

[0097] Corresponding to the above method embodiments, this application also provides an embodiment of a task execution timing prediction model training device. Figure 3 This diagram illustrates the structure of a task execution timing prediction model training device according to an embodiment of this application. Figure 3 As shown, the device includes:

[0098] The acquisition module 302 is configured to acquire historical task execution information of the target organization and construct a sample time series based on the historical task execution information;

[0099] The decomposition module 304 is configured to perform sequence decomposition on the sample time series to obtain the seasonal and trend terms associated with the sample time series.

[0100] The processing module 306 is configured to input the seasonal term into the initial prediction model and output first prediction information, input the trend term into the initial prediction model and output second prediction information, and input the sample time series into the initial prediction model to obtain third prediction information.

[0101] The construction module 308 is configured to construct seasonal loss, trend loss and overall loss based on the first prediction information, the second prediction information, the third prediction information and the sample time series;

[0102] Training module 310 is configured to construct a hybrid loss function based on the seasonal loss, the trend loss, and the overall loss, and to train the initial prediction model using the hybrid loss function to obtain the target prediction model.

[0103] In an optional embodiment, the processing module 306 is further configured to:

[0104] The initial prediction model includes an attention mechanism layer, wherein the influence factor associated with the attention mechanism layer is the task type of the task data contained in the historical task execution information; the seasonal item is feature extracted through the convolutional layer of the initial prediction model to obtain target seasonal features; the target seasonal features are dimensionality reduced through the pooling layer of the initial prediction model to obtain dimensionality-reduced seasonal features; the dimensionality-reduced seasonal features are processed through the attention mechanism layer, and the processing result is input into the fully connected layer of the initial prediction model for further processing to obtain the first prediction information; wherein, during the data processing of the initial prediction model, the dropout layer of the initial prediction model randomly discards neurons of the initial prediction model.

[0105] In an optional embodiment, the acquisition module 302 is further configured to:

[0106] The missing values ​​in the historical task execution information are queried and filled in using a preset interpolation algorithm. Under the assumption of normal distribution, based on the degree of variation of statistical data, the task data contained in the historical task execution information are sequentially judged to see if they are outliers. If so, the outliers are removed and filled in according to the interpolation algorithm. If not, the task data is retained until all the task data is processed to obtain sample task execution information. The sample time series is then constructed based on the sample task execution information.

[0107] In an optional embodiment, the building module 308 is further configured to:

[0108] The seasonal loss is constructed based on the first prediction information and the sample time series; the trend loss is constructed based on the second prediction information and the sample time series; and the overall loss is constructed based on the third prediction information and the sample time series.

[0109] In an optional embodiment, the training module 310 is further configured to:

[0110] Based on a preset first weighting coefficient, the seasonal loss and the trend loss are weighted and summed to obtain a component loss function; based on a preset second weighting coefficient, the component loss function and the overall loss are weighted and summed to obtain a mixed loss function.

[0111] In an optional embodiment, the training module 310 is further configured to:

[0112] The first weight coefficient and the second weight coefficient are optimized according to the preset min-max method to obtain the target mixture loss function; the initial prediction model is trained according to the target mixture loss function to obtain the target prediction model.

[0113] The task execution time-series prediction model training device provided in this application collects historical task execution information of the target organization and constructs a sample time series based on the historical task execution information; it performs sequence decomposition on the sample time series to obtain seasonal and trend terms associated with the sample time series; it inputs the seasonal terms into an initial prediction model to output first prediction information, inputs the trend terms into the initial prediction model to output second prediction information, and inputs the sample time series into the initial prediction model to obtain third prediction information; it constructs seasonal loss, trend loss, and overall loss based on the first, second, and third prediction information and the sample time series; it constructs a hybrid loss function based on the seasonal, trend, and overall losses, and trains the initial prediction model using the hybrid loss function to obtain a target prediction model. The obtained target prediction model achieves high-precision prediction of the target organization's task execution information without affecting computational efficiency, obtaining a balanced task execution plan.

[0114] The above is a schematic scheme of a task execution timing prediction model training device according to this embodiment. It should be noted that the technical solution of this task execution timing prediction model training device and the technical solution of the task execution timing prediction model training method described above belong to the same concept. Details not described in detail in the technical solution of the task execution timing prediction model training device can be found in the description of the technical solution of the task execution timing prediction model training method described above. Furthermore, the components in the device embodiment should be understood as functional modules necessary to implement each step of the program flow or each step of the method; these functional modules are not actual functional divisions or separations. The device claims defined by such a set of functional modules should be understood as a functional module architecture that primarily implements the solution through the computer program described in the specification, and not as a physical device that primarily implements the solution through hardware.

[0115] Figure 4 A structural block diagram of a computing device 400 according to an embodiment of this application is shown. The components of the computing device 400 include, but are not limited to, a memory 410 and a processor 420. The processor 420 is connected to the memory 410 via a bus 430, and a database 450 is used to store data.

[0116] The computing device 400 also includes an access device 440, which enables the computing device 400 to communicate via one or more networks 460. Examples of these networks include a Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the Internet. The access device 440 may include one or more of any type of wired or wireless network interface (e.g., a Network Interface Card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) interface, a Wi-MAX interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0117] In one embodiment of this application, the aforementioned components of the computing device 400 and Figure 4 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 4 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this application. Those skilled in the art can add or replace other components as needed.

[0118] The computing device 400 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or PCs. The computing device 400 can also be a mobile or stationary server.

[0119] The processor 420 is used to execute computer-executable instructions for each step of the task execution timing prediction model training method.

[0120] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the task execution time-series prediction model training method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the task execution time-series prediction model training method described above.

[0121] An embodiment of this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, are used to perform the steps of the task execution timing prediction model training method.

[0122] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the task execution timing prediction model training method described above. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the task execution timing prediction model training method described above.

[0123] An embodiment of this application also provides a chip that stores a computer program, which, when executed by the chip, implements the steps of the task execution timing prediction model training method.

[0124] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0126] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0127] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0128] The preferred embodiments disclosed above are merely illustrative of this application. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this application. These embodiments are selected and specifically described in this application to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. A method for training a task execution time-series prediction model, characterized in that, include: Collect historical task execution information of the target institution, and construct a sample time series based on the historical task execution information. The construction process of the sample time series is as follows: collect the number of cases filed by different courts and different types of cases according to a fixed time step, and summarize them into a time series dataset containing multiple variables. The sample time series is decomposed to obtain the seasonal and trend terms associated with the sample time series; The seasonal term is input into the initial prediction model to output first prediction information; the trend term is input into the initial prediction model to output second prediction information; and the sample time series is input into the initial prediction model to obtain third prediction information. The initial prediction model includes an attention mechanism layer, and the influence factor associated with the attention mechanism layer is the task type of the task data contained in the historical task execution information. Correspondingly, the first prediction information generation process involves extracting features from the seasonal term using the convolutional layer of the initial prediction model to obtain target seasonal features; reducing the dimensionality of the target seasonal features using the pooling layer of the initial prediction model to obtain dimensionality-reduced seasonal features; processing the dimensionality-reduced seasonal features using the attention mechanism layer; and inputting the processing result into the fully connected layer of the initial prediction model for further processing to obtain the first prediction information. During the data processing of the initial prediction model, the dropout layer of the initial prediction model randomly discards neurons of the initial prediction model. Based on the first prediction information, the second prediction information, the third prediction information, and the sample time series, seasonal loss, trend loss, and overall loss are constructed. Based on the seasonal loss, the trend loss, and the overall loss, a hybrid loss function is constructed, and the initial prediction model is trained using the hybrid loss function to obtain a target prediction model, wherein the target prediction model is used to predict the court case filing pattern. The hybrid loss function construction process is as follows: based on a preset first weight coefficient, the seasonal loss and the trend loss are weighted and summed to obtain a component loss function; based on a preset second weight coefficient, the component loss function and the overall loss are weighted and summed to obtain the hybrid loss function. Correspondingly, the process of training the initial prediction model to obtain the target prediction model is as follows: based on a preset min-max method, the first weight coefficient and the second weight coefficient are optimized to obtain a target hybrid loss function; based on the target hybrid loss function, the initial prediction model is trained to obtain the target prediction model.

2. The method according to claim 1, characterized in that, The step of constructing a sample time series based on the historical task execution information includes: Query the missing values ​​in the historical task execution information and fill in the missing values ​​using a preset interpolation algorithm; Under the assumption of normal distribution, based on the degree of variation in statistical data, we sequentially determine whether the task data contained in the historical task execution information are outliers. If so, remove the outlier and supplement the outlier according to the interpolation algorithm; If not, retain the task data until all task data has been processed to obtain sample task execution information, and construct the sample time series based on the sample task execution information.

3. The method according to claim 1, characterized in that, The step of constructing seasonal loss, trend loss, and overall loss based on the first prediction information, the second prediction information, the third prediction information, and the sample time series includes: The seasonal loss is constructed based on the first prediction information and the sample time series; The trend loss is constructed based on the second prediction information and the sample time series; The overall loss is constructed based on the third prediction information and the sample time series.

4. A training device for a task execution timing prediction model, characterized in that, include: The data acquisition module is configured to collect historical task execution information of the target institution and construct a sample time series based on the historical task execution information. The construction process of the sample time series is to collect the number of cases filed by different courts and different types of cases according to a fixed time step and summarize them into a time series dataset containing multiple variables. The decomposition module is configured to perform sequence decomposition on the sample time series to obtain the seasonal and trend terms associated with the sample time series. The processing module is configured to input the seasonal item into an initial prediction model and output first prediction information; input the trend item into the initial prediction model and output second prediction information; input the sample time series into the initial prediction model and obtain third prediction information. The initial prediction model includes an attention mechanism layer, and the influence factor associated with the attention mechanism layer is the task type of the task data contained in the historical task execution information. Correspondingly, the first prediction information generation process involves extracting features from the seasonal item through the convolutional layer of the initial prediction model to obtain target seasonal features; reducing the dimensionality of the target seasonal features through the pooling layer of the initial prediction model to obtain dimensionality-reduced seasonal features; processing the dimensionality-reduced seasonal features through the attention mechanism layer; and inputting the processing result into the fully connected layer of the initial prediction model for further processing to obtain the first prediction information. During the data processing of the initial prediction model, the dropout layer of the initial prediction model randomly discards neurons of the initial prediction model. The construction module is configured to construct seasonal loss, trend loss, and overall loss based on the first prediction information, the second prediction information, the third prediction information, and the sample time series. The training module is configured to construct a hybrid loss function based on the seasonal loss, the trend loss, and the overall loss, and to train the initial prediction model using the hybrid loss function to obtain a target prediction model, wherein the target prediction model is used to predict the pattern of court case filing. The hybrid loss function construction process is as follows: based on a preset first weight coefficient, the seasonal loss and the trend loss are weighted and summed to obtain a component loss function; based on a preset second weight coefficient, the component loss function and the overall loss are weighted and summed to obtain the hybrid loss function. Correspondingly, the process of training the initial prediction model to obtain the target prediction model is as follows: based on a preset min-max method, the first weight coefficient and the second weight coefficient are optimized to obtain a target hybrid loss function; based on the target hybrid loss function, the initial prediction model is trained to obtain the target prediction model.

5. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the steps of the task execution timing prediction model training method according to any one of claims 1 to 3.

6. A computer-readable storage medium storing computer instructions, characterized in that, When executed by the processor, this instruction implements the steps of the task execution timing prediction model training method according to any one of claims 1 to 3.

Images