Multi-flow interval prediction method and device based on model fusion

Abstract

The invention relates to a multi-flow interval prediction method and device based on model fusion, electronic equipment, a computer readable medium and a computer program product. The method comprises the following steps: dividing historical users into a plurality of different flow intervals, wherein the flow intervals at least comprise a high flow interval, a medium flow interval and a low flow interval; extracting historical users corresponding to two flow intervals one by one from the plurality of different flow intervals, and generating a plurality of training samples; respectively training a traffic prediction sub-model through the plurality of training samples; generating a multi-flow prediction model in a model fusion mode according to the plurality of trained flow prediction sub-models; and predicting the flow interval in which the current user is located through the multi-flow prediction model. The method can improve prediction accuracy, reduce errors, enhance model adaptability and robustness, improve traffic resource allocation efficiency, improve system data security and increase user satisfaction.

Description

Technical Field

[0001] This application relates to the field of computer information processing, and more specifically, to a method, apparatus, electronic device, computer-readable medium, and computer program product for multi-flow interval prediction based on model fusion. Background Technology

[0002] In many industries, accurate demand and resource forecasting is crucial for business decisions, especially in resource allocation, traffic management, and customer demand forecasting. Enterprises rely on extensive historical data and predictive models to anticipate resource consumption, customer behavior, or market trends across different demand ranges. This predictive capability helps companies allocate resources rationally amidst supply and demand fluctuations, optimize business costs, and improve customer experience. However, traditional forecasting methods often depend on a single model or a specific algorithm, making them susceptible to limitations imposed by data distribution, noise, and model structure. This impacts the accuracy and stability of forecast results, particularly posing significant challenges to forecast accuracy in the "medium-level" range.

[0003] Currently, the modeling methods commonly used in the industry include simple rule-based models and single machine learning models. For example, in scenarios involving income prediction or traffic allocation, some models divide users into three intervals—low, medium, and high—based on specific characteristics and model each interval separately. However, to ensure the stability and accuracy of the model's predictions, some data intervals are often excluded from the training samples, resulting in lower prediction accuracy for certain intervals (such as the medium income / traffic interval). For instance, some income prediction models / traffic allocation prediction models typically only use low income / traffic and high income / traffic data (i.e., data at opposite ends of two intervals, such as high-low or good-bad) to build the model, ignoring the data in the intermediate regions, such as neglecting to fully learn from the data in the medium income / medium traffic interval. This leads to bias in predicting medium-income users / medium traffic (level) and makes it difficult to meet actual needs.

[0004] Therefore, there is a need for a new method, device, electronic equipment, computer-readable medium, and computer program product for multi-flow interval prediction based on model fusion.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] In view of this, this application provides a multi-traffic interval prediction method, apparatus, electronic device, computer-readable medium, and computer program product based on model fusion, which can improve prediction accuracy, reduce errors, enhance model adaptability and robustness, improve traffic resource allocation efficiency, improve system data security, and increase user satisfaction.

[0007] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0008] According to one aspect of this application, a multi-traffic interval prediction method based on model fusion is proposed. The method includes: dividing historical users into multiple different traffic intervals, wherein the traffic intervals include at least a high traffic interval, a medium traffic interval, and a low traffic interval; extracting historical users corresponding to two traffic intervals from each of the multiple different traffic intervals to generate multiple training samples; training traffic prediction sub-models using the multiple training samples respectively; generating a multi-traffic prediction model by model fusion based on the trained multiple traffic prediction sub-models; and predicting the traffic interval in which the current user is located using the multi-traffic prediction model.

[0009] Optionally, historical users can be divided into multiple different traffic intervals, including: obtaining basic data of historical users; extracting traffic-related feature data from the basic data; and dividing historical users into multiple different traffic intervals based on the feature data.

[0010] Optionally, historical users can be divided into multiple different traffic intervals based on feature data, including: determining multiple different traffic intervals based on feature data and assigning corresponding traffic tags to each traffic interval; and assigning different traffic tags to historical users based on feature data.

[0011] Optionally, when the traffic range includes high traffic range, medium traffic range, and low traffic range, historical users corresponding to two traffic ranges are extracted one by one from multiple different traffic ranges to generate multiple training samples, including: generating high-medium training samples from historical users corresponding to high traffic range and medium traffic range; generating low-medium training samples from historical users corresponding to low traffic range and medium traffic range; and generating high-low training samples from historical users corresponding to high traffic range and low traffic range.

[0012] Optionally, the traffic prediction sub-model is trained using the multiple training samples, including: constructing an initial traffic prediction model using a binary classification model; training the initial traffic prediction model using the multiple training samples, and generating the traffic prediction sub-model after training is completed.

[0013] Optionally, an initial traffic prediction model is constructed using a binary classification model, including: constructing an initial traffic prediction model using XGBoost; and / or constructing an initial traffic prediction model using LightGBM; and / or constructing an initial traffic prediction model using CatBoost.

[0014] Optionally, a multi-traffic prediction model is generated by model fusion based on multiple trained traffic prediction sub-models, including: generating the multi-traffic prediction model by weighted averaging based on multiple trained traffic prediction sub-models; and / or generating the multi-traffic prediction model by voting based on multiple trained traffic prediction sub-models; and / or generating the multi-traffic prediction model by neural network based on multiple trained traffic prediction sub-models.

[0015] Optionally, the multi-traffic prediction model is generated from the trained multi-traffic prediction sub-models using a neural network, including: generating the calculation results of the trained multi-traffic prediction sub-models; and inputting the multi-calculation results as new features into the neural network model for retraining to generate the multi-traffic prediction model.

[0016] Optionally, predicting the current user's traffic range using the multi-traffic prediction model includes: inputting the current user's information into the multi-traffic prediction model; the multi-traffic prediction model calculating the user information to give the probability of the user being in each traffic range; and outputting the traffic range corresponding to the highest probability as the result.

[0017] According to one aspect of this application, a multi-traffic interval prediction device based on model fusion is proposed. The device includes: an interval module for dividing historical users into multiple different traffic intervals, wherein the traffic intervals include at least a high traffic interval, a medium traffic interval, and a low traffic interval; a sample module for extracting historical users corresponding to two traffic intervals from the multiple different traffic intervals one by one to generate multiple training samples; a training module for training traffic prediction sub-models using the multiple training samples respectively; a model module for generating a multi-traffic prediction model based on the trained multiple traffic prediction sub-models through model fusion; and a prediction module for predicting the traffic interval in which the current user is located using the multi-traffic prediction model.

[0018] According to one aspect of this application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the method as described above.

[0019] According to one aspect of this application, a computer-readable medium is provided having a computer program stored thereon that, when executed by a processor, implements the method described above.

[0020] According to one aspect of this application, a computer program product is provided, comprising: a computer program / instructions that, when executed by a processor, implement the method as described above.

[0021] According to the multi-traffic interval prediction method, apparatus, electronic device, computer-readable medium, and computer program product based on model fusion of this application, historical users are divided into multiple different traffic intervals, which include at least high traffic intervals, medium traffic intervals, and low traffic intervals; historical users corresponding to two traffic intervals are extracted one by one from the multiple different traffic intervals to generate multiple training samples; traffic prediction sub-models are trained using the multiple training samples respectively; a multi-traffic prediction model is generated by model fusion based on the trained multiple traffic prediction sub-models; the method of predicting the traffic interval of the current user using the multi-traffic prediction model can improve prediction accuracy, reduce errors, enhance model adaptability and robustness, improve traffic resource allocation efficiency, improve system data security, and increase user satisfaction.

[0022] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0023] The above and other objects, features, and advantages of this application will become more apparent from the detailed description of exemplary embodiments with reference to the accompanying drawings. The drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0024] Figure 1 This is a flowchart illustrating a multi-flow interval prediction method based on model fusion according to an exemplary embodiment.

[0025] Figure 2 This is a flowchart illustrating a multi-flow interval prediction method based on model fusion according to another exemplary embodiment.

[0026] Figure 3 This is a flowchart illustrating a multi-flow interval prediction method based on model fusion according to another exemplary embodiment.

[0027] Figure 4 This is a block diagram illustrating a multi-flow interval prediction device based on model fusion according to an exemplary embodiment.

[0028] Figure 5 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0029] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, they are provided so that this application will be thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0030] Figure 1 This is a flowchart illustrating a multi-flow interval prediction method based on model fusion according to an exemplary embodiment. The multi-flow interval prediction method 10 based on model fusion includes at least steps S102 to S110.

[0031] like Figure 1 As shown, in S102, historical users are divided into multiple different traffic intervals, which include at least a high traffic interval, a medium traffic interval, and a low traffic interval. For example, basic data of historical users can be obtained; traffic-related feature data can be extracted from the basic data; and historical users can be divided into multiple different traffic intervals based on the feature data.

[0032] In specific scenarios, historical user data can be extracted from databases or data warehouses, including but not limited to user historical access records, usage duration, access frequency, and other traffic-related characteristic data. Basic data typically includes user account information, access behavior, historical transactions, and may also involve user device usage, network connection type, etc. Ensure that the basic data is cleaned and anomalies are removed to guarantee the accuracy and consistency when segmenting traffic zones.

[0033] This includes dividing historical users into multiple different traffic intervals based on feature data, including: determining multiple different traffic intervals based on feature data and assigning corresponding traffic tags to each traffic interval; and assigning different traffic tags to historical users based on feature data.

[0034] In specific scenarios, traffic-related features are extracted from the basic data, such as daily average traffic, monthly average traffic, and peak traffic per visit. These features reflect users' traffic usage behavior. Using feature engineering methods, basic statistics (mean, median, maximum, minimum, etc.), volatility (such as standard deviation), and specific user usage patterns (such as peak traffic during specific time periods) can be extracted from the traffic data.

[0035] Based on the extracted feature data, users can be divided into different traffic ranges using quantiles, clustering algorithms (such as K-means clustering), or based on business rules. For example, the top 20% of users can be defined as the high-traffic range, the bottom 20% as the low-traffic range, and the rest as the medium-traffic range.

[0036] Assign labels to each traffic range: Assign labels (e.g., "high", "medium", "low") to the high traffic range, medium traffic range, and low traffic range respectively. These labels will be used for subsequent training sample generation and model training.

[0037] In S104, historical users corresponding to two traffic intervals are extracted one by one from multiple different traffic intervals to generate multiple training samples. Specifically, the process of extracting historical users corresponding to two traffic intervals from multiple different traffic intervals to generate multiple training samples includes: generating high-medium training samples from historical users corresponding to high and medium traffic intervals; generating low-medium training samples from historical users corresponding to low and medium traffic intervals; and generating high-low training samples from historical users corresponding to high and low traffic intervals.

[0038] When more intervals are included, such as four traffic intervals (high, medium, low, and inactive), combined samples such as "high-medium," "medium-low," "high-low," "high-inactive," "medium-inactive," and "low-inactive" can be generated. By increasing the types of combined samples, the model's ability to capture the interaction features of different intervals is improved. In particular, the combination of four intervals can further refine the feature information and improve the model's predictive ability in medium or other special traffic intervals.

[0039] In step S106, the traffic prediction sub-model is trained using the multiple training samples. An initial traffic prediction model can be constructed using a binary classification model; the initial traffic prediction model is then trained using the multiple training samples, and the traffic prediction sub-model is generated after training is complete.

[0040] A binary classification model can be chosen as the initial model for traffic prediction. Algorithms suitable for classification tasks, such as Support Vector Machine (SVM), Logistic Regression, Decision Tree, and XGBoost, can be used. The choice of model should be based on the complexity of the features of samples in different intervals and the interpretability of the model in practical applications.

[0041] More specifically, the generated high-medium, low-medium, and high-low training samples can be used to train three different traffic prediction sub-models. Each sub-model focuses on learning the feature differences of specific interval combinations, improving the model's ability to distinguish between different intervals. During training, these sub-models optimize their parameters through multiple iterations to ensure the model's prediction accuracy across each interval.

[0042] Cross-validation can also be used to evaluate the performance of sub-models. By adjusting the model's hyperparameters (such as tree depth and regularization coefficient), optimal model performance can be obtained. During training, the performance of each model on data in different intervals is recorded to facilitate reasonable weight allocation in subsequent model fusion.

[0043] In S108, a multi-traffic prediction model is generated by fusing multiple trained traffic prediction sub-models. For each trained sub-model, a suitable model fusion method is selected.

[0044] For example, the multi-flow prediction model can be generated by weighted averaging multiple trained flow prediction sub-models; based on the prediction performance of each sub-model in different flow ranges, corresponding weights are assigned, and the prediction results of the sub-models are weighted and averaged. This method is simple, efficient, and computationally inexpensive, and is suitable for situations where the performance of each model varies.

[0045] Alternatively, for example, a multi-traffic prediction model can be generated by voting on multiple trained traffic prediction sub-models; a voting mechanism is then used to select the traffic interval with the highest frequency as the final prediction result based on the prediction results of each sub-model. The voting method is suitable for situations where the prediction results of each model are similar, and is appropriate for improving the robustness of the model.

[0046] Alternatively, the multi-traffic prediction model can be generated using a neural network based on multiple trained traffic prediction sub-models: by constructing a simple neural network layer, the prediction results of each sub-model are input into the neural network and trained, enabling the neural network to automatically learn and adjust the weights of each model based on the input, thereby further improving the accuracy of the fusion model.

[0047] The fusion model was debugged and optimized, and the impact of different fusion methods on the prediction results was analyzed to select the optimal fusion method. Multiple experiments were conducted for each method, and the effectiveness of the fusion model was evaluated using a validation set. The fusion method with the highest prediction accuracy across multiple intervals was selected as the final model.

[0048] In S110, the traffic range in which the current user is located is predicted using the multi-traffic prediction model. For example, the user information of the current user is input into the multi-traffic prediction model; the multi-traffic prediction model calculates the probability of the user being in each traffic range based on the user information; and the traffic range corresponding to the highest probability is output as the result.

[0049] According to the multi-traffic interval prediction method based on model fusion of this application, historical users are divided into multiple different traffic intervals, which include at least high traffic intervals, medium traffic intervals, and low traffic intervals. Historical users corresponding to two traffic intervals are extracted one by one from each of the multiple different traffic intervals to generate multiple training samples. Traffic prediction sub-models are trained using these multiple training samples. A multi-traffic prediction model is generated by model fusion based on the trained multiple traffic prediction sub-models. This method of predicting the current user's traffic interval using the multi-traffic prediction model can improve prediction accuracy, reduce errors, enhance model adaptability and robustness, improve traffic resource allocation efficiency, improve system data security, and increase user satisfaction.

[0050] It should be clearly understood that this application describes how specific examples are formed and used, but the principles of this application are not limited to any details of these examples. Rather, based on the teachings of the disclosure of this application, these principles can be applied to many other embodiments.

[0051] Figure 2 This is a flowchart illustrating a multi-flow interval prediction method based on model fusion according to another exemplary embodiment. Figure 2 The process shown in step 20 is... Figure 1 The flowchart shown describes step S108, "Generate a multi-traffic prediction model by model fusion based on the trained multi-traffic prediction sub-models."

[0052] like Figure 2 As shown in step S202, training data is input into multiple trained traffic prediction sub-models, generating calculation results for each. Sample data is selected from the training set and preprocessed to ensure consistency in format and features between the training data and those used during sub-model training. This data includes user traffic-related features such as historical access volume, access frequency, and device usage characteristics.

[0053] Multiple traffic prediction sub-models are identified. These sub-models may be generated from the same type of base model or belong to different model series. This application is not limited to this.

[0054] The training data is fed into each sub-model for prediction. Each sub-model outputs different prediction results based on its specific training samples (such as training samples in high-medium, low-medium, and high-low intervals). The output of each sub-model is the probability of the traffic flow interval predicted by that model or the specific predicted value.

[0055] The output of each sub-model is typically a vector representing the probabilities of its corresponding two traffic intervals. For example, for a sample with inputs of high and medium intervals, the sub-model output could be [P(High), P(Medium)]; similarly, for a sample with inputs of low and medium intervals, the sub-model output could be [P(Low), P(Medium)]. The outputs of all sub-models are recorded and used as new features to provide input for subsequent neural network models. Each sample ultimately generates multiple sub-model predictions as its new feature set.

[0056] In S204, multiple calculation results are used as new features and input into the neural network model for retraining. To handle these new features, a multi-layer neural network can be constructed; a common network structure might include an input layer, several hidden layers, and an output layer. The input layer receives feature vectors generated by the sub-model, the hidden layers capture the non-linear relationships between features, and the output layer is used for the final flow range prediction. The appropriate number of layers and neurons can be selected based on data complexity and requirements. Activation functions such as ReLU can be used in the network to introduce non-linear relationships, and regularization and other methods can be employed to prevent overfitting.

[0057] The outputs of the aforementioned sub-models are used as new feature data and input into the neural network model for training. During training, the real traffic interval labels corresponding to each sample are used as supervision signals, and the backpropagation algorithm is used to adjust the network parameters, gradually optimizing the model's weights and biases. During training, hyperparameters such as the learning rate and batch size are adjusted to ensure that the model effectively learns patterns between features and labels.

[0058] The training effect of the neural network model is evaluated using a validation set. Precision, accuracy, and recall are calculated to observe whether the model's predictive ability for each traffic interval meets expectations. If the model performs poorly in certain intervals, the network structure or sample weights can be adjusted to improve the overall performance of the model.

[0059] In S206, the multi-traffic prediction model is generated when the loss function during training meets a threshold. A suitable loss function for traffic interval prediction, such as cross-entropy loss or mean squared error (MSE), is selected to measure the error between the model's prediction and the true label. The threshold for the loss function is set according to business requirements; that is, when the model's loss value drops to this threshold or does not change significantly over multiple iterations, the model can be considered to have reached ideal accuracy.

[0060] When the loss function meets the set threshold condition or the training reaches the predetermined maximum number of rounds, training stops and the parameters of the current neural network model are saved as the final multi-traffic prediction model. This model has integrated the prediction information of multiple sub-models and further optimized through neural networks, achieving efficient prediction of user traffic ranges.

[0061] Figure 3 This is a flowchart illustrating a multi-flow interval prediction method based on model fusion according to another exemplary embodiment. Figure 3 The process shown in step 30 is... Figure 1 The process shown in the diagram is described in detail in step S110, which involves "predicting the current user's traffic range using the multi-traffic prediction model".

[0062] like Figure 3 As shown, in S302, the user information of the current user is input into the multi-traffic prediction model. For example, traffic-related feature data of the current user can be collected and preprocessed into a format that meets the input requirements of the multi-traffic prediction model. Input features may include the user's recent access volume, time frequency, network bandwidth, etc., and these data should be consistent with the features of the training samples to ensure the accuracy of the model's predictions.

[0063] In S304, the multi-traffic prediction model calculates the probability of the user being located in each traffic interval based on the user information. For example, the current user's feature data can be input into the multi-traffic prediction model, and the model calculates the probability of the current user belonging to each traffic interval based on the learning results of each sub-model. For instance, the probabilities of high, medium, and low traffic intervals are 0.6, 0.3, and 0.1, respectively, indicating that the user is most likely to belong to the high traffic interval.

[0064] In S306, the traffic interval corresponding to the highest probability is output as the result. For example, the interval corresponding to the highest probability can be selected from the interval probabilities output by the model as the prediction result for the current user. This result is returned to the business system for further decision support in resource allocation or traffic management.

[0065] Those skilled in the art will understand that all or part of the steps of the above embodiments are implemented as a computer program executed by a CPU. When the computer program is executed by the CPU, it performs the functions defined by the method provided in this application. The program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disk.

[0066] Furthermore, it should be noted that the above figures are merely illustrative representations of the processes included in the method according to exemplary embodiments of this application, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0067] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0068] Figure 4 This is a block diagram illustrating a multi-flow interval prediction device based on model fusion according to an exemplary embodiment. Figure 4 As shown, the multi-flow interval prediction device 40 based on model fusion includes: interval module 402, sample module 404, training module 406, model module 408, and prediction module 410.

[0069] The interval module 402 is used to divide historical users into multiple different traffic intervals, the traffic intervals including at least a high traffic interval, a medium traffic interval, and a low traffic interval; the interval module 402 is also used to obtain basic data of historical users; extract traffic-related feature data from the basic data; and divide historical users into multiple different traffic intervals according to the feature data.

[0070] The sample module 404 is used to extract historical users corresponding to two of the multiple different traffic intervals one by one to generate multiple training samples; the sample module 404 is also used to generate high and medium training samples through historical users corresponding to high and medium traffic intervals; generate low and medium training samples through historical users corresponding to low and medium traffic intervals; and generate high and low training samples through historical users corresponding to high and low traffic intervals.

[0071] The training module 406 is used to train the traffic prediction sub-model using the multiple training samples respectively; the training module 406 is also used to construct an initial traffic prediction model using a binary classification model; the initial traffic prediction model is trained using the multiple training samples respectively, and the traffic prediction sub-model is generated after training is completed.

[0072] Model module 408 is used to generate a multi-traffic prediction model by model fusion based on multiple trained traffic prediction sub-models; model module 408 is also used to generate the multi-traffic prediction model by weighted averaging based on multiple trained traffic prediction sub-models; model module 408 is also used to generate the multi-traffic prediction model by voting based on multiple trained traffic prediction sub-models; model module 408 is also used to generate the multi-traffic prediction model by neural network based on multiple trained traffic prediction sub-models.

[0073] The prediction module 410 is used to predict the current user's current traffic range using the multi-traffic prediction model. The prediction module 410 is also used to input the current user's information into the multi-traffic prediction model; the multi-traffic prediction model calculates the probability of the user being in each traffic range based on the user information; and outputs the traffic range corresponding to the highest probability as the result.

[0074] According to the multi-traffic interval prediction device based on model fusion of this application, historical users are divided into multiple different traffic intervals, including at least high traffic intervals, medium traffic intervals, and low traffic intervals; historical users corresponding to two traffic intervals are extracted one by one from each of the multiple different traffic intervals to generate multiple training samples; traffic prediction sub-models are trained using the multiple training samples respectively; and a multi-traffic prediction model is generated by model fusion based on the trained multiple traffic prediction sub-models; the method of predicting the traffic interval of the current user using the multi-traffic prediction model can improve prediction accuracy, reduce errors, enhance model adaptability and robustness, improve traffic resource allocation efficiency, improve system data security, and increase user satisfaction.

[0075] like Figure 5 As shown, this application provides an electronic device including a processor 510, a memory 520, and a bus, wherein the processor 510 and the memory 520 communicate with each other through the bus 540.

[0076] Memory 520 is used to store computer programs;

[0077] When the processor 510 executes the program stored in the memory 520, it implements the multi-flow interval prediction method based on model fusion of any of the above embodiments.

[0078] Communication interface 520 is used for communication between the above-mentioned electronic device and other devices.

[0079] The memory 520 may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory 520 may also be at least one storage device located remotely from the aforementioned processor 510.

[0080] If the methods described in this application are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms.

[0081] This application provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the multi-traffic interval prediction method based on model fusion of any of the above embodiments. For example, historical users can be divided into multiple different traffic intervals, including at least high-traffic intervals, medium-traffic intervals, and low-traffic intervals; historical users corresponding to two traffic intervals can be extracted one by one from each of the multiple different traffic intervals to generate multiple training samples; traffic prediction sub-models can be trained using the multiple training samples respectively; a multi-traffic prediction model can be generated by model fusion based on the trained multiple traffic prediction sub-models; and the multi-traffic prediction model can be used to predict the traffic interval in which the current user is located.

[0082] Exemplary embodiments of this application have been specifically shown and described above. It should be understood that this application is not limited to the detailed structures, arrangements, or implementation methods described herein; rather, this application is intended to cover various modifications and equivalent arrangements contained within the spirit and scope of the appended claims.

Claims

1. A multi-flow interval prediction method based on model fusion, characterized in that, include: Historical users are divided into multiple different traffic ranges, and the traffic ranges include at least a high traffic range, a medium traffic range, and a low traffic range. Historical users corresponding to two traffic intervals are extracted one by one from multiple different traffic intervals to generate multiple training samples; The traffic prediction sub-model is trained using the multiple training samples respectively; A multi-traffic prediction model is generated by fusion of multiple trained traffic prediction sub-models. The multi-traffic prediction model is used to predict the current traffic range of the user.

2. The method as described in claim 1, characterized in that, Historical users are divided into several different traffic ranges, including: Obtain basic data of historical users; Traffic-related feature data are extracted from the basic data; Historical users are divided into multiple different traffic ranges based on feature data.

3. The method as described in claim 2, characterized in that, Based on characteristic data, historical users are divided into several different traffic intervals, including: Based on the feature data, multiple different traffic intervals are determined, and corresponding traffic labels are assigned to each traffic interval. Different traffic tags are assigned to historical users based on feature data.

4. The method as described in claim 1, characterized in that, When the traffic range includes high traffic range, medium traffic range, and low traffic range. Historical users corresponding to two traffic intervals are extracted one by one from multiple different traffic intervals to generate multiple training samples, including: High school training samples are generated by using historical users corresponding to high and medium traffic ranges. Low- and medium-traffic training samples are generated from historical users corresponding to low-traffic and medium-traffic ranges. High and low training samples are generated by using historical users corresponding to high and low traffic ranges.

5. The method as described in claim 1, characterized in that, The traffic prediction sub-model is trained using the multiple training samples, including: An initial model for traffic prediction is constructed using a binary classification model; The initial traffic prediction model is trained using the multiple training samples, and the traffic prediction sub-model is generated after training is completed.

6. The method as described in claim 5, characterized in that, An initial traffic prediction model is constructed using a binary classification model, including: Build an initial traffic prediction model using XGBoost; and / or An initial traffic prediction model was constructed using LightGBM; and / or An initial model for traffic prediction was built using CatBoost.

7. The method as described in claim 1, characterized in that, A multi-traffic prediction model is generated by fusing multiple trained traffic prediction sub-models, including: The multi-flow prediction model is generated by weighted averaging multiple trained flow prediction sub-models; and / or The multi-flow prediction model is generated by voting based on multiple trained flow prediction sub-models; and / or The multi-traffic prediction model is generated using a neural network based on multiple trained traffic prediction sub-models.

8. The method as described in claim 7, characterized in that, The multi-traffic prediction model is generated using a neural network based on multiple trained traffic prediction sub-models, including: Generate the calculation results of multiple trained traffic prediction sub-models; The multiple calculation results are used as new features and input into the neural network model for retraining to generate the multi-flow prediction model.

9. The method as described in claim 1, characterized in that, The multi-traffic prediction model is used to predict the current user's traffic range, including: Input the current user's user information into the multi-traffic prediction model; The multi-traffic prediction model calculates the user information and gives the probability that the user is located in each traffic interval. The traffic range corresponding to the highest probability is output as the result.

10. A multi-flow interval prediction device based on model fusion, characterized in that, include: The interval module is used to divide historical users into multiple different traffic intervals, which include at least a high traffic interval, a medium traffic interval, and a low traffic interval. The sample module is used to extract historical users corresponding to two of the multiple different traffic intervals one by one, and generate multiple training samples. The training module is used to train the traffic prediction sub-model using the multiple training samples respectively; The model module is used to generate a multi-traffic prediction model by model fusion based on multiple trained traffic prediction sub-models. The prediction module is used to predict the current user's traffic range using the multi-traffic prediction model.

11. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 9.

12. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 9.

13. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 9.

Images