Decision engine data processing method and device based on artificial intelligence

Through artificial intelligence technology, the decision data of the decision engine is processed, and the rules are predicted and updated, which solves the problem of difficulty in updating the decision engine when business changes, and improves the efficiency and accuracy of the decision engine.

CN118966288APending Publication Date: 2024-11-15紫金诚征信有限公司
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411091386.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Existing decision-making engines need to frequently update and maintain decision-making rules when the business environment changes, resulting in wasted time and resources, and poor reusability of code writing, which cannot meet the enterprise's accurate and timely requirements for decision-making systems.

Method used

Using the decision engine data processing method based on artificial intelligence, collect historical decision data, use preset static covariate encoder and time-varying variable encoder to determine static and dynamic features, input deep time series models for feature selection and fusion, and predict decisions. rules and update decision engine rules through hot deployment technology.

Benefits of technology

It improves the efficiency and accuracy of the decision-making engine, reduces the timeliness of rule updates, and improves the timeliness and accuracy of corporate decision-making.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118966288A_ABST
    Figure CN118966288A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a decision engine data processing method and device based on artificial intelligence, and the method comprises the steps: collecting historical decision data, determining corresponding static features and dynamic features through a preset static covariable encoder and a preset time-varying variable encoder, carrying out the effective feature screening operation, and carrying out the effective feature screening operation; inputting a feature fusion layer of the preset depth time sequence model, determining a corresponding comprehensive feature, determining the depth time sequence model according to the comprehensive feature, a preset real feature and a back propagation algorithm, obtaining latest trend data to be decided, and outputting the latest trend data to be decided. And inputting the latest trend data to be decided into the depth time sequence model for model prediction, determining a corresponding prediction decision rule, obtaining a corresponding decision result according to the decision engine updating rule, and feeding back the decision result to a user interface. According to the invention, the efficiency and accuracy of the decision engine can be improved based on decision prediction and hot deployment technologies.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing, and specifically to a decision engine data processing method and device based on artificial intelligence. Background Art

[0002] A decision engine is a software system that can handle complex decision logic, including conditional judgments, loops, branches, etc. Its main function is to help organizations or individuals automate the decision-making process. In the decision engine platform, customers will write code according to customized rule requirements. After the code is written, it will be integrated into the decision engine to automatically execute the decision rules, reduce manual intervention, and improve the speed and consistency of decision-making. Decision engines have been widely used in insurance, finance, e-commerce and other industries.

[0003] Although the application of decision engines can greatly reduce human intervention and improve decision-making speed, the decision rules used by decision engines are still based on a system of predefined rules, which are usually formulated by business experts or domain experts based on their knowledge and experience.

[0004] However, as the business environment changes, decision rules need to be constantly updated and maintained, which may require a lot of time and resources. Analysts are also required to write code based on their understanding of corporate rules and product trends. The reusability of the written code is poor and cannot meet the company's requirements for accurate and timely decision-making systems. Summary of the invention

[0005] In response to the problems in the prior art, the present application provides a decision engine data processing method and device based on artificial intelligence, which can improve the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0006] In order to solve at least one of the above problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a decision engine data processing method based on artificial intelligence, comprising:

[0008] Collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of a preset deep time series model, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, input the static features and the dynamic features respectively into the variable selection network of the preset deep time series model, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm;

[0009] Acquire the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rules;

[0010] Determine whether the prediction decision rule is consistent with the rule in the decision engine. If not, update the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule, determine the corresponding decision engine update rule, obtain the corresponding decision result according to the decision engine update rule, and feed back the decision result to the user interface.

[0011] Further, the determining of the corresponding pre-processed decision data set according to the historical decision data and the preset data processing rules includes:

[0012] Determine corresponding seasonal trend characteristics according to the historical decision data and the preset deep neural network model;

[0013] Decision data features are extracted from the historical decision data, and time alignment is performed on the decision data features and the seasonal trend features to determine a corresponding preprocessed decision data set.

[0014] Furthermore, the corresponding static features and dynamic features are determined respectively by using a preset static covariate encoder and a preset time-varying variable encoder, including:

[0015] Performing a static feature encoding operation on the preprocessed decision data set through a preset fully connected neural network layer to determine the corresponding static features;

[0016] A dependency capturing operation is performed on the preprocessed decision data set through a preset conversion layer to determine corresponding dynamic features, wherein the dynamic features include past dynamic features and future dynamic features.

[0017] Furthermore, the step of inputting the static features and the dynamic features into the variable selection network of the preset deep time series model to perform an effective feature screening operation includes:

[0018] Inputting the static features into the first variable selection network of the preset deep time series model, performing effective feature screening operations, and determining corresponding static effective features;

[0019] Inputting the past dynamic features in the dynamic features into the second variable selection network of the preset deep time series model, performing an effective feature screening operation, and determining the corresponding past dynamic effective features;

[0020] Inputting the future dynamic features in the dynamic features into the third variable selection network of the preset deep time series model, performing an effective feature screening operation, and determining the corresponding future dynamic effective features;

[0021] The corresponding dynamic effective features are determined according to the past dynamic effective features and the future dynamic effective features.

[0022] Furthermore, the static effective features and dynamic effective features obtained after the effective feature screening operation are input into the feature fusion layer of the preset deep time series model to determine the corresponding comprehensive features, including:

[0023] Input the static effective features and dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, perform feature fusion according to the preset multimodal fusion rules, and determine the corresponding fusion features;

[0024] A time-step dependency capture operation is performed based on the fusion features and a preset multi-head attention algorithm to determine the corresponding comprehensive features.

[0025] Furthermore, determining the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm includes:

[0026] Determine the corresponding prediction feature according to the comprehensive feature and the preset time fusion decoder;

[0027] Determine a corresponding loss function according to the predicted features and the preset real features;

[0028] The parameters of the static covariate encoder, the time-varying variable encoder and the variable selection network are updated according to the minimum loss function and the back-propagation algorithm to determine the corresponding deep time series model.

[0029] Further, the determining whether the prediction decision rule is consistent with the rule in the decision engine, and if not, updating the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule to determine the corresponding decision engine update rule, includes:

[0030] Determine whether the prediction decision rule is consistent with the rule in the decision engine, and if not, compile the prediction decision rule according to a preset compiler to determine the corresponding compiled file;

[0031] The compiled file is input into a preset class loader, and the rules in the decision engine are updated according to the preset class loader to determine the corresponding decision engine update rules.

[0032] In a second aspect, the present application provides a decision engine data processing device based on artificial intelligence, comprising:

[0033] A decision rule model determination module is used to collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of a preset deep time series model, respectively determine corresponding static features and dynamic features through a preset static covariate encoder and a preset time-varying variable encoder, respectively input the static features and the dynamic features into the variable selection network of the preset deep time series model, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, preset real features and a back propagation algorithm;

[0034] A decision rule prediction module is used to obtain the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule;

[0035] The decision engine update module is used to determine whether the prediction decision rules are consistent with the rules in the decision engine. If not, the rules in the decision engine are updated according to the prediction decision rules and preset hot deployment rules, the corresponding decision engine update rules are determined, the corresponding decision results are obtained according to the decision engine update rules, and the decision results are fed back to the user interface.

[0036] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the artificial intelligence-based decision engine data processing method when executing the program.

[0037] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the artificial intelligence-based decision engine data processing method.

[0038] In a fifth aspect, the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the artificial intelligence-based decision engine data processing method.

[0039] It can be seen from the above technical scheme that the present application provides a decision engine data processing method and device based on artificial intelligence, which collects historical decision data, determines the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, performs effective feature screening operations, inputs the feature fusion layer of the preset deep time series model, determines the corresponding comprehensive features, determines the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtains the latest trend data to be decided, inputs the latest trend data to be decided into the deep time series model for model prediction, determines the corresponding prediction decision rules, obtains the corresponding decision results according to the decision engine update rules, and feeds back the decision results to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 This is one of the flowcharts of the artificial intelligence-based decision engine data processing method in the embodiment of the present application;

[0042] Figure 2 This is a second flow chart of the decision engine data processing method based on artificial intelligence in an embodiment of the present application;

[0043] Figure 3 The third flowchart of the decision engine data processing method based on artificial intelligence in the embodiment of the present application;

[0044] Figure 4 This is a fourth flow chart of the decision engine data processing method based on artificial intelligence in an embodiment of the present application;

[0045] Figure 5This is a fifth flow chart of the decision engine data processing method based on artificial intelligence in an embodiment of the present application;

[0046] Figure 6 This is a sixth flow chart of the artificial intelligence-based decision engine data processing method in the embodiment of the present application;

[0047] Figure 7 FIG7 is a flowchart of a decision engine data processing method based on artificial intelligence in an embodiment of the present application;

[0048] Figure 8 This is a structural diagram of an artificial intelligence-based decision engine data processing device in an embodiment of the present application;

[0049] Fig. 9 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application.

[0050] Reference numerals:

[0051] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0054] Considering that as the business environment changes, the decision engine needs to continuously update and maintain decision rules, which may require a lot of time and resources, and analysts are also required to write code based on the understanding of enterprise rules and product trends. The reusability of the written code is poor, and it cannot meet the company's requirements for accurate and timely decision systems. The present application provides a decision engine data processing method and device based on artificial intelligence, which collects historical decision data, determines the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, performs effective feature screening operations, inputs the feature fusion layer of the preset deep time series model, determines the corresponding comprehensive features, determines the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtains the latest trend data to be decided, inputs the latest trend data to be decided into the deep time series model for model prediction, determines the corresponding prediction decision rules, obtains the corresponding decision results according to the decision engine update rules, and feeds the decision results back to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0055] In order to improve the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology, the present application provides an embodiment of a decision engine data processing method based on artificial intelligence, see Figure 1 , the decision engine data processing method based on artificial intelligence specifically includes the following contents:

[0056] Step S101: collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of the preset deep time series model, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, input the static features and the dynamic features into the variable selection network of the preset deep time series model respectively, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm;

[0057] Optionally, in this embodiment, the purpose of this step is to train a deep time series model, which can discover the relationship between historical decision rules and time patterns, and make predictions of decision rules based on historical information and future trends, laying the foundation for the next step of hot loading the predicted decision rules into the decision engine.

[0058] Optionally, in the present embodiment, the technical means for implementing this step is to divide the historical decision data into static variables and dynamic variables, perform network training separately, and perform feature fusion to obtain a deep time series model.

[0059] It is understandable that designing networks for different data sources in multi-step prediction tasks will improve the model and help users identify globally important features, temporal patterns, and important events.

[0060] Optionally, in this embodiment, static covariates: variables that do not change over time, such as store locations;

[0061] Time-dependent Inputs: Variables that change over time;

[0062] Past-observed inputs: The past is known but the future is unknown, such as historical passenger flow.

[0063] Apriori-known Future Inputs: Both the past and the future are known, such as holidays.

[0064] Optionally, in this embodiment, historical decision data is collected, and a corresponding pre-processed decision data set is determined according to the historical decision data and preset data processing rules.

[0065] Specifically, historical decision data includes decision rules made by users for various application scenarios in the past, where the application scenarios include seasonal trends, weather, holidays, etc.

[0066] The preset data processing rules are the rules for obtaining decision features based on historical decision data and seasonal laws. First, the historical decision data is input into the pre-trained deep neural network, and the neural network model is used to analyze the time features in the historical decision data to identify the seasonal trend features of the decision data, such as holidays, seasonal changes, etc. Secondly, sales-related features are extracted from the historical decision data, such as price, promotion, inventory level, customer evaluation, etc. Then, the decision data features are aligned with the seasonal trend features to ensure that they are on the same time series for joint analysis, where time alignment is the process of matching and synchronizing time series data from different data sources according to a unified time reference. Parse the date and time information in the time series data into a unified time format. Determine the size of the aligned time window, for example, select daily, weekly, or monthly as a time window. If data is missing at certain time points, interpolation methods (such as linear interpolation, polynomial interpolation, nearest neighbor interpolation, etc.) can be used to estimate the missing values.

[0067] In time alignment, a common time unit should be selected as the basis for all time series. For data that does not need to be accurate to every second, "day" can be selected as the basis.

[0068] It can be understood that in the above step, seasonality means that the mean of some periods is different from the mean of other periods. This problem causes the series to be non-stationary. This is why it is important to analyze seasonality during data preprocessing.

[0069] Optionally, in this embodiment, the preprocessed decision data set is input into the input layer of a preset deep time series model, and the corresponding static features and dynamic features are respectively determined by a preset static covariate encoder and a preset time-varying variable encoder.

[0070] Specifically, Static Covariate Encoder: The Static Covariate Encoder is a fully connected neural network layer that encodes static features into a fixed-size vector that can then be used by other parts of the model. Use the Static Covariate Encoder to process invariant features such as store location, product type, product name, etc.

[0071] Dynamic Covariate Encoder: The Dynamic Covariate Encoder captures long-term dependencies in time series data through transformation layers. Time-varying variables are features that change over time, for example, sales or temperature per day.

[0072] Among them, time-varying variables are further divided into:

[0073] Past dynamic features: such as sales data in recent weeks, reflecting historical sales trends.

[0074] Future dynamic features: predict sales data for the next few weeks and reflect potential sales trends.

[0075] It can be understood that by separately identifying and processing static features and dynamic features, it helps to improve the interpretability of the model.

[0076] Optionally, in this embodiment, the static features and the dynamic features are respectively input into the variable selection network of the preset deep time series model to perform an effective feature screening operation.

[0077] Specifically, for static inputs (such as user gender, product category, etc., which do not change over time), a dedicated VSN (variable selection network) is used to learn the importance of these static features. The parameters of the static VSN are independent in the entire model and are not shared with other VSNs.

[0078] For dynamic input, it is further divided into past dynamic features and future dynamic features, which are also learned using dedicated VSNs.

[0079] For historical time series data (past dynamic features), another VSN is used to select the past time points most relevant to the prediction task. This VSN focuses on capturing dynamic changes and temporal dependencies in time series data.

[0080] For future known information (future dynamic features, such as holidays and seasons at a specific time point), a third VSN is used to process this information. This VSN helps the model understand the potential impact of future events on the prediction results.

[0081] Each VSN learns the importance of its own input type through an independent parameter learning process. This means that each VSN has its own set of parameters that are updated independently during training.

[0082] It is understandable that static features and dynamic features use different variable selection networks, which can help the model identify the most important features in different time dimensions.

[0083] Specifically, Variable Selection Networks (VSNs) are used to determine which features are most important for the prediction task.

[0084] VSN consists of multiple gating units (such as sigmoid or softmax functions), each of which corresponds to an input feature. The output of the gating unit determines the activity of the corresponding feature in subsequent processing. At the same time, by using the attention mechanism to learn the correlation between features, VSN can adaptively focus on the features that are more important to the current prediction task.

[0085] VSN learns the importance weight of each feature through the training process. The weight indicates the contribution of each feature to the prediction result. VSN can dynamically adjust the importance of features to adapt to the variability of time series data.

[0086] It can be understood that in the above steps, VSN considers the interaction between features and identifies which feature combinations are most effective for prediction. In this way, effective features are screened to obtain effective static features and dynamic features, laying the foundation for subsequent model training.

[0087] Optionally, in this embodiment, the static effective features and dynamic effective features obtained after the effective feature screening operation are input into the feature fusion layer of the preset deep time series model to determine the corresponding comprehensive features.

[0088] Specifically, the features after training are multimodally fused to obtain fused features for subsequent calculation and update of loss functions. Multimodal fusion refers to combining data modalities from different data sources or with different characteristics to obtain a more comprehensive feature representation. This fusion method can enhance the model's ability to understand and process information and improve the performance of prediction and classification tasks.

[0089] Multimodal fusion uses maximum activation pooling to combine information from static and dynamic features. By selectively integrating features, the interpretability of the model can be improved because the decision-making process of the model is more focused on key features.

[0090] A multi-head self-attention mechanism is applied to the fused features to further capture the dependencies between time steps and obtain comprehensive features. Each head of the multi-head self-attention mechanism independently calculates the attention weight, which can be regarded as processing the feature representations of different subspaces in parallel, while learning the representation of input data from multiple perspectives. Each head may capture different dependencies, and the outputs of all heads are merged to form a comprehensive representation.

[0091] It is understandable that multi-head self-attention can better capture long-term and short-term dependencies in time series. Different attention heads can learn different representation subspaces to provide the model with richer feature information. At the same time, the model can adaptively focus on the most important time steps for the current task and improve the accuracy of prediction.

[0092] By applying a multi-head self-attention mechanism on the fused features, the model can make full use of the information of static and dynamic features to capture complex patterns and trends in time series data, thereby improving the prediction performance.

[0093] Optionally, in this embodiment, a deep time series model is determined based on the comprehensive features, preset real features and a back propagation algorithm.

[0094] Specifically, the temporal fusion decoder uses this information to generate predictions based on the comprehensive features after applying attention. The temporal fusion decoder uses a sequence-to-sequence architecture to convert the output of the encoder into a prediction sequence. The decoder can combine the fused representation of static covariates and time-varying variables to predict future values.

[0095] The predicted features output by the decoder are compared with the preset real features, the difference between the predicted results and the actual values ​​is measured, the corresponding loss function is calculated, and the loss function is optimized. The model parameters are updated by backpropagation of the minimized loss function to obtain a trained deep time series model.

[0096] It can be understood that through the above steps, a deep time series model is trained, which can generate corresponding decision rules based on the data features in the existing decision-making data, helping enterprises to make better corporate decisions based on the existing situation.

[0097] Step S102: Acquire the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule;

[0098] Optionally, in this embodiment, the data to be decided is input into a trained deep time series model for model prediction. The model can make predictions based on seasonal characteristics, static data characteristics, and dynamic data characteristics in the data to be decided. For example:

[0099] Suppose a clothing retailer wants to predict sales for the upcoming winter season and automatically generate sales decision rules.

[0100] First, collect historical decision data, including sales data from the past few years, including sales volume, price, promotional activities and seasonal events. Preprocess according to preset rules, determine seasonal trend characteristics, extract decision data features, and align the two in time so that they can be jointly analyzed later.

[0101] Secondly, the preprocessed data is fed into the preset deep time series model for model training. The fully connected layer is used to extract static features such as brand and product category, while the conversion layer is used to capture the dynamic changes in sales data. The most important features for sales forecasting are screened out through the variable selection network, and these features are fused and further represented by comprehensive changes. The predicted features output by the decoder are compared with the actual features, and the model parameters are updated in reverse. The update process is repeated to obtain a trained deep time series model.

[0102] Finally, by receiving the latest decision-making data input by users, the trained deep time series model can predict sales trends in the next few months and help retailers formulate corresponding inventory and marketing strategies.

[0103] for example:

[0104] Inventory management: If the forecast shows that sales of a certain product will increase in winter, increase the inventory of that product in advance.

[0105] Promotional Events: When the winter sales trough is predicted, plan promotional events to stimulate demand.

[0106] Pricing strategy: Adjust pricing strategy based on forecasted sales volume and price sensitivity.

[0107] New Product Launch: Arrange the launch of new products when high demand for them is predicted in the market.

[0108] Step S103: Determine whether the prediction decision rule is consistent with the rule in the decision engine. If not, update the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule, determine the corresponding decision engine update rule, obtain the corresponding decision result according to the decision engine update rule, and feed back the decision result to the user interface.

[0109] Optionally, in this embodiment, the decision engine is a system that executes business logic and decisions, which records the decisions originally used to execute strategies, for example, when the season enters spring, reduce the inventory of winter clothing while increasing the inventory ratio of summer clothing.

[0110] Optionally, in this embodiment, it is determined whether the prediction decision rule and the rule in the decision engine are consistent. If not, the rule in the decision engine is updated according to the prediction decision rule and the preset hot deployment rule to determine the corresponding decision engine update rule.

[0111] Optionally, in this embodiment, the purpose of this step is to obtain the predicted decision rules through the model and determine whether the predicted rules are consistent with the original rules. If they are inconsistent, hot deployment technology is used to directly change the rules in the decision engine for application, which can greatly reduce the time of rule application and improve the timeliness of enterprise decision-making.

[0112] Optionally, in this embodiment, hot deployment can be achieved in the decision engine scenario, where the rule input automatically takes effect without the intervention of development and operation and maintenance, and the steps are as follows:

[0113] 1. Compile the java code and use the compilation capability of tools.jar (tool library) in jdk (Java Development Kit) to compile the java file and finally form a class file.

[0114] 2. Use the class loader of SpringBoot (open source Java framework) to load it into JVM (Java virtual machine): Use the class loader LaunchedURLClassLoader (launched URL class loader) in SpringBoot to load the class file into JVM.

[0115] 3. Load the code logic of this decision calculation into the rule configuration logic flow of the decision engine. It will take effect.

[0116] It can be understood that the above steps solve the problem of repeated compilation and release of code in decision calculations, allowing Java code to take effect immediately without restarting the service, thereby improving the flexibility and responsiveness of the decision engine, reducing manpower investment and reducing costs.

[0117] Optionally, in this embodiment, the decision engine receives real-time decision rules output by the model, and loads the rules into the decision engine for execution based on hot deployment, obtains decision results based on the latest rules, and displays the decision results on the user interface, which can help users quickly make adaptive decisions based on the latest seasonal data and trends, thereby improving user experience and business competitiveness.

[0118] This example shows how to train a time series model in this embodiment to generate decision rules in real time, and load the decision rules into the decision engine for decision making based on hot deployment, thereby helping users improve their competitiveness.

[0119] From the above description, it can be seen that the artificial intelligence-based decision engine data processing method provided in the embodiment of the present application can collect historical decision data, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, perform effective feature screening operations, input the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtain the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, determine the corresponding prediction decision rules, obtain the corresponding decision results according to the decision engine update rules, and feed back the decision results to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0120] In one embodiment of the decision engine data processing method based on artificial intelligence of the present application, see Figure 2 , and can also include the following:

[0121] Step S201: determining corresponding seasonal trend characteristics according to the historical decision data and a preset deep neural network model;

[0122] Step S202: extracting decision data features from the historical decision data, performing a time alignment operation on the decision data features and the seasonal trend features, and determining a corresponding preprocessed decision data set.

[0123] Optionally, in this embodiment, the historical decision data includes decision rules formulated by users for various application scenarios in the past, where the application scenarios include seasonal trends, weather, holidays, etc.

[0124] Optionally, in this embodiment, the preset data processing rules are rules for obtaining decision features based on historical decision data and seasonal laws. First, the historical decision data is input into a pre-trained deep neural network, and the time features in the historical decision data are analyzed using the neural network model to identify the seasonal trend features of the decision data, such as holidays, seasonal changes, etc. Secondly, sales-related features such as price, promotion, inventory level, customer evaluation, etc. are extracted from the historical decision data. Then, the decision data features are aligned with the seasonal trend features to ensure that they are on the same time series for joint analysis, where time alignment is the process of matching and synchronizing time series data from different data sources according to a unified time reference. Parse the date and time information in the time series data into a unified time format. Determine the size of the aligned time window, for example, select daily, weekly or monthly as a time window. If data is missing at certain time points, interpolation methods (such as linear interpolation, polynomial interpolation, nearest neighbor interpolation, etc.) can be used to estimate the missing values.

[0125] In time alignment, a common time unit should be selected as the basis for all time series. For data that does not need to be accurate to every second, "day" can be selected as the basis.

[0126] It can be understood that in the above step, seasonality means that the mean of some periods is different from the mean of other periods. This problem causes the series to be non-stationary. This is why it is important to analyze seasonality during data preprocessing.

[0127] In the preprocessing stage, the features in the dataset are classified into seasonal trend features and other features. This step helps the model to better understand the different patterns in the data.

[0128] Through step S202, this embodiment obtains a preprocessed data set, which integrates seasonal trends and can help the model better understand data patterns and discover the laws of seasonal factors in subsequent model training.

[0129] In one embodiment of the decision engine data processing method based on artificial intelligence of the present application, see Figure 3 , and can also include the following:

[0130] Step S301: performing a static feature encoding operation on the preprocessed decision data set through a preset fully connected neural network layer to determine corresponding static features;

[0131] Step S302: performing a dependency capture operation on the preprocessed decision data set through a preset conversion layer to determine corresponding dynamic features, wherein the dynamic features include past dynamic features and future dynamic features.

[0132] Optionally, in this embodiment, a static covariate encoder: A static covariate encoder is a fully connected neural network layer that encodes static features into a fixed-size vector that can then be used by other parts of the model. Use a static covariate encoder to process invariant features such as store location, product type, product name, etc.

[0133] Dynamic Covariate Encoder: The Dynamic Covariate Encoder captures long-term dependencies in time series data through transformation layers. Time-varying variables are features that change over time, for example, sales or temperature per day.

[0134] Among them, time-varying variables (dynamic features) are further divided into:

[0135] Past dynamic features: such as sales data in recent weeks, reflecting historical sales trends.

[0136] Future dynamic features: predict sales data for the next few weeks and reflect potential sales trends.

[0137] It can be understood that by separately identifying and processing static features and dynamic features, it helps to improve the interpretability of the model.

[0138] Through step S302, this embodiment successfully identifies static variables and dynamic variables through the static encoder and the dynamic encoder, laying a foundation for subsequent models to train different variables separately and improve data interpretability.

[0139] In one embodiment of the decision engine data processing method based on artificial intelligence of the present application, see Figure 4 , and can also include the following:

[0140] Step S401: inputting the static features into the first variable selection network of the preset deep time series model, performing effective feature screening operations, and determining corresponding static effective features;

[0141] Step S402: inputting the past dynamic features in the dynamic features into the second variable selection network of the preset deep time series model, performing an effective feature screening operation, and determining the corresponding past dynamic effective features;

[0142] Step S403: inputting the future dynamic features in the dynamic features into the third variable selection network of the preset deep time series model, performing an effective feature screening operation, and determining the corresponding future dynamic effective features;

[0143] Step S404: Determine the corresponding dynamic valid features according to the past dynamic valid features and the future dynamic valid features.

[0144] Optionally, in this embodiment, for static inputs (such as user gender, product category, etc., which do not change over time), a dedicated VSN (variable selection network) is used to learn the importance of these static features. The parameters of the static VSN are independent in the entire model and are not shared with other VSNs.

[0145] For dynamic input, it is further divided into past dynamic features and future dynamic features, which are also learned using dedicated VSNs.

[0146] For historical time series data (past dynamic features), another VSN is used to select the past time points most relevant to the prediction task. This VSN focuses on capturing dynamic changes and temporal dependencies in time series data.

[0147] For future known information (future dynamic features, such as holidays and seasons at a specific time point), a third VSN is used to process this information. This VSN helps the model understand the potential impact of future events on the prediction results.

[0148] Each VSN learns the importance of its own input type through an independent parameter learning process. This means that each VSN has its own set of parameters that are updated independently during training.

[0149] It is understandable that static features and dynamic features use different variable selection networks, which can help the model identify the most important features in different time dimensions.

[0150] Specifically, Variable Selection Networks (VSNs) are used to determine which features are most important for the prediction task.

[0151] VSN consists of multiple gating units (such as sigmoid or softmax functions), each of which corresponds to an input feature. The output of the gating unit determines the activity of the corresponding feature in subsequent processing. At the same time, by using the attention mechanism to learn the correlation between features, VSN can adaptively focus on the features that are more important to the current prediction task.

[0152] VSN learns the importance weight of each feature through the training process. The weight indicates the contribution of each feature to the prediction result. VSN can dynamically adjust the importance of features to adapt to the variability of time series data.

[0153] It can be understood that in the above steps, VSN considers the interaction between features and identifies which feature combinations are most effective for prediction. In this way, effective features are screened to obtain effective static features and dynamic features, laying the foundation for subsequent model training.

[0154] Through step S404, this embodiment successfully performs further effective feature screening on static features and dynamic features, obtains static effective features and dynamic effective features, and lays a foundation for subsequent feature fusion.

[0155] In one embodiment of the decision engine data processing method based on artificial intelligence of the present application, see Figure 5 , and can also include the following:

[0156] Step S501: input the static effective features and dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, perform feature fusion according to the preset multimodal fusion rules, and determine the corresponding fusion features;

[0157] Step S502: Perform a time-step dependency capture operation based on the fusion features and a preset multi-head attention algorithm to determine the corresponding comprehensive features.

[0158] Optionally, in this embodiment, the features after training are multimodally fused to obtain fused features for subsequent calculation and update of loss functions. Multimodal fusion refers to combining data modalities from different data sources or with different characteristics to obtain a more comprehensive feature representation. This fusion method can enhance the model's ability to understand and process information and improve the performance of prediction and classification tasks.

[0159] Multimodal fusion uses maximum activation pooling to combine information from static and dynamic features. By selectively integrating features, the interpretability of the model can be improved because the decision-making process of the model is more focused on key features.

[0160] A multi-head self-attention mechanism is applied to the fused features to further capture the dependencies between time steps and obtain comprehensive features. Each head of the multi-head self-attention mechanism independently calculates the attention weight, which can be regarded as processing the feature representations of different subspaces in parallel, while learning the representation of input data from multiple perspectives. Each head may capture different dependencies, and the outputs of all heads are merged to form a comprehensive representation.

[0161] It is understandable that multi-head self-attention can better capture long-term and short-term dependencies in time series. Different attention heads can learn different representation subspaces to provide the model with richer feature information. At the same time, the model can adaptively focus on the most important time steps for the current task and improve the accuracy of prediction.

[0162] By applying a multi-head self-attention mechanism on the fused features, the model can make full use of the information of static and dynamic features to capture complex patterns and trends in time series data, thereby improving the prediction performance.

[0163] Through step S502, this embodiment successfully obtains comprehensive features that integrate dynamic features and static features and adapt to the attention mechanism, laying the foundation for subsequent updating of the loss function.

[0164] In one embodiment of the decision engine data processing method based on artificial intelligence of the present application, see Figure 6 , and can also include the following:

[0165] Step S601: Determine corresponding prediction features according to the comprehensive features and a preset time fusion decoder;

[0166] Step S602: determining a corresponding loss function according to the predicted feature and the preset real feature;

[0167] Step S603: performing parameter update operations on the static covariate encoder, the time-varying variable encoder, and the variable selection network according to the minimum loss function and the back propagation algorithm to determine the corresponding deep time series model.

[0168] Optionally, in this embodiment, the temporal fusion decoder uses this information to generate predictions based on the comprehensive features after applying attention. The temporal fusion decoder uses a sequence-to-sequence architecture to convert the output of the encoder into a prediction sequence. The decoder can combine the fused representation of static covariates and time-varying variables to predict future values.

[0169] The predicted features output by the decoder are compared with the preset real features, the difference between the predicted results and the actual values ​​is measured, the corresponding loss function is calculated, and the loss function is optimized. The model parameters (static covariate encoder, time-varying variable encoder, and variable selection network parameters) are updated by back propagation of the minimized loss function to obtain a trained deep time series model.

[0170] It can be understood that through the above steps, a deep time series model is trained, which can generate corresponding decision rules based on the data features in the existing decision-making data, helping enterprises to make better corporate decisions based on the existing situation.

[0171] Through step S602, this embodiment successfully obtains a trained deep time series model for predicting decision rules, helping enterprises to make decisions more effectively and timely.

[0172] In one embodiment of the decision engine data processing method based on artificial intelligence of the present application, see Figure 7 , and can also include the following:

[0173] Step S701: determining whether the prediction decision rule is consistent with the rule in the decision engine; if not, compiling the prediction decision rule according to a preset compiler to determine a corresponding compilation file;

[0174] Step S702: input the compiled file into a preset class loader, perform an update operation on the rules in the decision engine according to the preset class loader, and determine the corresponding decision engine update rules.

[0175] Optionally, in this embodiment, the decision engine is a system that executes business logic and decisions, which records the decisions originally used to execute strategies, for example, when the season enters spring, reduce the inventory of winter clothing while increasing the inventory ratio of summer clothing.

[0176] Optionally, in this embodiment, it is determined whether the prediction decision rule and the rule in the decision engine are consistent. If not, the rule in the decision engine is updated according to the prediction decision rule and the preset hot deployment rule to determine the corresponding decision engine update rule.

[0177] Optionally, in this embodiment, the purpose of this step is to obtain the predicted decision rules through the model and determine whether the predicted rules are consistent with the original rules. If they are inconsistent, hot deployment technology is used to directly change the rules in the decision engine for application, which can greatly reduce the time of rule application and improve the timeliness of enterprise decision-making.

[0178] Optionally, in this embodiment, hot deployment can be achieved in the decision engine scenario, where the rule input automatically takes effect without the intervention of development and operation and maintenance, and the steps are as follows:

[0179] 1. Compile the java code and use the compilation capability of tools.jar (tool library) in jdk (Java Development Kit) to compile the java file and finally form a class file.

[0180] 2. Use the class loader of SpringBoot (open source Java framework) to load it into JVM (Java virtual machine): Use the class loader LaunchedURLClassLoader (launched URL class loader) in SpringBoot to load the class file into JVM.

[0181] 3. Load the code logic of this decision calculation into the rule configuration logic flow of the decision engine. It will take effect.

[0182] It can be understood that the above steps solve the problem of repeated compilation and release of code in decision calculations, allowing Java code to take effect immediately without restarting the service, thereby improving the flexibility and responsiveness of the decision engine, reducing manpower investment and reducing costs.

[0183] Optionally, in this embodiment, the decision engine receives real-time decision rules output by the model, and loads the rules into the decision engine for execution based on hot deployment, obtains decision results based on the latest rules, and displays the decision results on the user interface, which can help users quickly make adaptive decisions based on the latest seasonal data and trends, thereby improving user experience and business competitiveness.

[0184] Through step S702, this embodiment successfully loads the model prediction results into the decision engine in real time through hot deployment technology, which can effectively save the time of decision engine application decision, improve the flexibility and response speed of the decision engine, and lay the foundation for improving the competitiveness of the enterprise.

[0185] In order to improve the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology, the present application provides an embodiment of an artificial intelligence-based decision engine data processing device for implementing all or part of the artificial intelligence-based decision engine data processing method, see Figure 8 , the decision engine data processing device based on artificial intelligence specifically includes the following contents:

[0186] A decision rule model determination module 10 is used to collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of a preset deep time series model, respectively determine corresponding static features and dynamic features through a preset static covariate encoder and a preset time-varying variable encoder, respectively input the static features and the dynamic features into the variable selection network of the preset deep time series model, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm;

[0187] A decision rule prediction module 20 is used to obtain the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule;

[0188] The decision engine update module 30 is used to determine whether the prediction decision rules are consistent with the rules in the decision engine. If not, the rules in the decision engine are updated according to the prediction decision rules and the preset hot deployment rules, the corresponding decision engine update rules are determined, the corresponding decision results are obtained according to the decision engine update rules, and the decision results are fed back to the user interface.

[0189] From the above description, it can be seen that the artificial intelligence-based decision engine data processing device provided in the embodiment of the present application can collect historical decision data, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, perform effective feature screening operations, input the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtain the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, determine the corresponding prediction decision rules, obtain the corresponding decision results according to the decision engine update rules, and feed back the decision results to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0190] From the hardware level, in order to improve the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology, the present application provides an embodiment of an electronic device for implementing all or part of the content of the artificial intelligence-based decision engine data processing method, and the electronic device specifically includes the following content:

[0191] Processor, memory, communication interface and bus; wherein the processor, memory and communication interface communicate with each other through the bus; the communication interface is used to realize information transmission between the decision engine data processing method based on artificial intelligence and related devices such as core business systems, user terminals and related databases; the logic controller can be a desktop computer, a tablet computer and a mobile terminal, etc., but the present embodiment is not limited thereto. In the present embodiment, the logic controller can be implemented with reference to the embodiment of the decision engine data processing method based on artificial intelligence and the embodiment of the decision engine data processing method based on artificial intelligence in the embodiment, and the contents thereof are incorporated herein, and the repeated parts are not repeated.

[0192] It is understandable that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0193] In practical applications, part of the decision engine data processing method based on artificial intelligence can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific selection can be based on the processing capability of the client device and the limitations of the user's usage scenario. This application does not limit this. If all operations are completed in the client device, the client device may also include a processor.

[0194] The client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0195] Fig. 9 FIG. 9 is a schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Fig. 9 As shown, the electronic device 9600 may include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. It is worth noting that Fig. 9 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0196] In one embodiment, the function of the decision engine data processing method based on artificial intelligence can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0197] Step S101: collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of the preset deep time series model, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, input the static features and the dynamic features into the variable selection network of the preset deep time series model respectively, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm;

[0198] Step S102: Acquire the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule;

[0199] Step S103: Determine whether the prediction decision rule is consistent with the rule in the decision engine. If not, update the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule, determine the corresponding decision engine update rule, obtain the corresponding decision result according to the decision engine update rule, and feed back the decision result to the user interface.

[0200] From the above description, it can be seen that the electronic device provided in the embodiment of the present application collects historical decision data, determines the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, performs an effective feature screening operation, inputs the feature fusion layer of the preset deep time series model, determines the corresponding comprehensive features, determines the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtains the latest trend data to be decided, inputs the latest trend data to be decided into the deep time series model for model prediction, determines the corresponding prediction decision rules, obtains the corresponding decision results according to the decision engine update rules, and feeds back the decision results to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0201] In another embodiment, the artificial intelligence-based decision engine data processing method can be configured separately from the central processing unit 9100. For example, the artificial intelligence-based decision engine data processing method can be configured as a chip connected to the central processing unit 9100, and the functions of the artificial intelligence-based decision engine data processing method can be realized through the control of the central processing unit.

[0202] like Fig. 9 As shown, the electronic device 9600 may also include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Fig. 9 In addition, the electronic device 9600 may also include Fig. 9 For components not shown, reference may be made to the prior art.

[0203] like Fig. 9 As shown, the central processing unit 9100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0204] The memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 9100 may execute the program stored in the memory 9140 to implement information storage or processing, etc.

[0205] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0206] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 9600 through the central processor 9100.

[0207] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0208] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0209] Based on different communication technologies, multiple communication modules 9110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module, etc. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, thereby realizing a common telecommunication function. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processor 9100, so that recording can be performed on the local machine through the microphone 9132, and the sound stored on the local machine can be played through the speaker 9131.

[0210] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all the steps of the decision engine data processing method based on artificial intelligence in the above embodiments, wherein the execution subject is a server or a client. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, all the steps of the decision engine data processing method based on artificial intelligence in the above embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0211] Step S101: collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of the preset deep time series model, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, input the static features and the dynamic features into the variable selection network of the preset deep time series model respectively, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm;

[0212] Step S102: Acquire the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule;

[0213] Step S103: Determine whether the prediction decision rule is consistent with the rule in the decision engine. If not, update the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule, determine the corresponding decision engine update rule, obtain the corresponding decision result according to the decision engine update rule, and feed back the decision result to the user interface.

[0214] From the above description, it can be seen that the computer-readable storage medium provided in the embodiment of the present application collects historical decision data, determines the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, performs an effective feature screening operation, inputs the feature fusion layer of the preset deep time series model, determines the corresponding comprehensive features, determines the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtains the latest trend data to be decided, inputs the latest trend data to be decided into the deep time series model for model prediction, determines the corresponding prediction decision rules, obtains the corresponding decision results according to the decision engine update rules, and feeds back the decision results to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0215] The embodiments of the present application also provide a computer program product capable of implementing all the steps of the artificial intelligence-based decision engine data processing method in the above embodiments, where the execution subject is a server or a client. When the computer program / instruction is executed by a processor, the steps of the artificial intelligence-based decision engine data processing method are implemented. For example, the computer program / instruction implements the following steps:

[0216] Step S101: collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of the preset deep time series model, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, input the static features and the dynamic features into the variable selection network of the preset deep time series model respectively, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm;

[0217] Step S102: Acquire the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule;

[0218] Step S103: Determine whether the prediction decision rule is consistent with the rule in the decision engine. If not, update the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule, determine the corresponding decision engine update rule, obtain the corresponding decision result according to the decision engine update rule, and feed back the decision result to the user interface.

[0219] From the above description, it can be seen that the computer program product provided in the embodiment of the present application collects historical decision data, determines the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, performs an effective feature screening operation, inputs the feature fusion layer of the preset deep time series model, determines the corresponding comprehensive features, determines the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm, obtains the latest trend data to be decided, inputs the latest trend data to be decided into the deep time series model for model prediction, determines the corresponding prediction decision rules, obtains the corresponding decision results according to the decision engine update rules, and feeds back the decision results to the user interface, thereby improving the efficiency and accuracy of the decision engine based on decision prediction and hot deployment technology.

[0220] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0221] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0222] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0223] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0224] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A decision engine data processing method based on artificial intelligence, It is characterized in that The method comprises: Collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of a preset deep time series model, determine the corresponding static features and dynamic features respectively through a preset static covariate encoder and a preset time-varying variable encoder, input the static features and the dynamic features respectively into the variable selection network of the preset deep time series model, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm; Acquire the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rules; Determine whether the prediction decision rule is consistent with the rule in the decision engine. If not, update the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule, determine the corresponding decision engine update rule, obtain the corresponding decision result according to the decision engine update rule, and feed back the decision result to the user interface.

2. The decision engine data processing method based on artificial intelligence according to claim 1, It is characterized in that The determining of the corresponding pre-processed decision data set according to the historical decision data and the preset data processing rules includes: Determine corresponding seasonal trend characteristics according to the historical decision data and the preset deep neural network model; Decision data features are extracted from the historical decision data, and time alignment is performed on the decision data features and the seasonal trend features to determine a corresponding preprocessed decision data set.

3. The decision engine data processing method based on artificial intelligence according to claim 1, It is characterized in that The corresponding static features and dynamic features are determined respectively by presetting a static covariate encoder and a presetting a time-varying variable encoder, including: Performing a static feature encoding operation on the preprocessed decision data set through a preset fully connected neural network layer to determine the corresponding static features; A dependency capturing operation is performed on the preprocessed decision data set through a preset conversion layer to determine corresponding dynamic features, wherein the dynamic features include past dynamic features and future dynamic features.

4. The decision engine data processing method based on artificial intelligence according to claim 1, It is characterized in that The step of inputting the static features and the dynamic features into the variable selection network of the preset deep time series model to perform an effective feature screening operation includes: Inputting the static features into the first variable selection network of the preset deep time series model, performing effective feature screening operations, and determining corresponding static effective features; Inputting the past dynamic features in the dynamic features into the second variable selection network of the preset deep time series model, performing an effective feature screening operation, and determining the corresponding past dynamic effective features; Inputting the future dynamic features in the dynamic features into the third variable selection network of the preset deep time series model, performing an effective feature screening operation, and determining the corresponding future dynamic effective features; The corresponding dynamic effective features are determined according to the past dynamic effective features and the future dynamic effective features.

5. The decision engine data processing method based on artificial intelligence according to claim 1, It is characterized in that The static effective features and dynamic effective features obtained after the effective feature screening operation are input into the feature fusion layer of the preset deep time series model to determine the corresponding comprehensive features, including: Input the static effective features and dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, perform feature fusion according to the preset multimodal fusion rules, and determine the corresponding fusion features; A time-step dependency capture operation is performed based on the fusion features and a preset multi-head attention algorithm to determine the corresponding comprehensive features.

6. The decision engine data processing method based on artificial intelligence according to claim 1, It is characterized in that Determining the deep time series model according to the comprehensive features, the preset real features and the back propagation algorithm includes: Determine the corresponding prediction feature according to the comprehensive feature and the preset time fusion decoder; Determine a corresponding loss function according to the predicted features and the preset real features; The parameters of the static covariate encoder, the time-varying variable encoder and the variable selection network are updated according to the minimum loss function and the back-propagation algorithm to determine the corresponding deep time series model.

7. The decision engine data processing method based on artificial intelligence according to claim 1, It is characterized in that The determining whether the prediction decision rule is consistent with the rule in the decision engine, and if not, updating the rule in the decision engine according to the prediction decision rule and the preset hot deployment rule to determine the corresponding decision engine update rule, includes: Determine whether the prediction decision rule is consistent with the rule in the decision engine, and if not, compile the prediction decision rule according to a preset compiler to determine the corresponding compiled file; The compiled file is input into a preset class loader, and the rules in the decision engine are updated according to the preset class loader to determine the corresponding decision engine update rules.

8. A decision engine data processing device based on artificial intelligence, It is characterized in that The device comprises: A decision rule model determination module is used to collect historical decision data, determine a corresponding preprocessed decision data set according to the historical decision data and preset data processing rules, input the preprocessed decision data set into the input layer of a preset deep time series model, respectively determine corresponding static features and dynamic features through a preset static covariate encoder and a preset time-varying variable encoder, respectively input the static features and the dynamic features into the variable selection network of the preset deep time series model, perform an effective feature screening operation, input the static effective features and the dynamic effective features obtained after the effective feature screening operation into the feature fusion layer of the preset deep time series model, determine the corresponding comprehensive features, and determine the deep time series model according to the comprehensive features, preset real features and a back propagation algorithm; A decision rule prediction module is used to obtain the latest trend data to be decided, input the latest trend data to be decided into the deep time series model for model prediction, and determine the corresponding prediction decision rule; The decision engine update module is used to determine whether the prediction decision rules are consistent with the rules in the decision engine. If not, the rules in the decision engine are updated according to the prediction decision rules and preset hot deployment rules, the corresponding decision engine update rules are determined, the corresponding decision results are obtained according to the decision engine update rules, and the decision results are fed back to the user interface.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the steps of the artificial intelligence-based decision engine data processing method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the artificial intelligence-based decision engine data processing method described in any one of claims 1 to 7 are implemented.

Citation Information

Cited By

  • Dynamic network system maintenance system and method under confrontation condition

    CN119629071A