User traffic prediction method and device, computer device and storage medium
By filtering and adjusting the user traffic prediction network and combining data from multiple influencing factors, the problem of low accuracy caused by relying on historical traffic flow in traditional methods has been solved, and more accurate bank branch customer traffic prediction has been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INDUSTRIAL AND COMMERCIAL BANK OF CHINA
- Filing Date
- 2022-11-23
- Publication Date
- 2026-06-23
AI Technical Summary
Traditional methods for predicting customer traffic rely on historical traffic data, resulting in low accuracy and an inability to accurately predict customer traffic at bank branches, leading to excessively long waiting times for customers or a waste of human resources.
By acquiring data on influencing factors from multiple categories, filtering the data on influencing factors for the target category, and adjusting the initial user traffic prediction network (excluding the target influencing factor data), predictions are made by comprehensively considering multiple influencing factor data using self-attention networks and fully connected neural networks.
It improves the accuracy of user traffic prediction, avoids over-reliance on data from a single influencing factor, and enhances the accuracy of prediction results.
Smart Images
Figure CN115915233B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of big data technology, and in particular to a user traffic prediction method, apparatus, computer equipment, and storage medium. Background Technology
[0002] Bank services and product promotions are conducted through bank branches, which require a certain number of tellers to handle the daily needs of customers. However, the number of customers is dynamic; too few tellers can lead to long waiting times and negatively impact user experience, while too many can result in unnecessary waste of human resources. Therefore, accurately predicting customer traffic at bank branches is essential.
[0003] Traditional user traffic prediction methods utilize prediction networks such as Recurrent Neural Networks (RNNs), taking historical pedestrian traffic as input data and weather factors as an influencing factor to predict user traffic during a target period. However, this method relies too heavily on historical pedestrian traffic data, resulting in low accuracy in the predicted user traffic data. Summary of the Invention
[0004] Therefore, it is necessary to provide a user traffic prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product to address the aforementioned technical problems.
[0005] Firstly, this application provides a user traffic prediction method. The method includes:
[0006] Acquire data on influencing factors across multiple categories; the influencing factor data refers to the data corresponding to the influencing factors of user traffic within historical time periods.
[0007] In the influencing factor data of each category, the target influencing factor data of the target category is filtered, and the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data is adjusted to obtain the user traffic prediction network corresponding to each influencing factor data.
[0008] For each category of influencing factor data, based on the user traffic prediction network corresponding to the influencing factor data of that category, predict the user traffic data for the target time period corresponding to the influencing factor data of that category;
[0009] Based on the user traffic data for the target time period corresponding to each of the aforementioned influencing factors, determine the target user traffic data for the target time period.
[0010] Optionally, after obtaining the influencing factor data for multiple categories, the method further includes:
[0011] Obtain the data type of each type of influencing factor data, and select the user traffic prediction network that matches the data type of each type of influencing factor data from among the user traffic prediction networks, and use it as the user traffic prediction network corresponding to each type of influencing factor data.
[0012] Optionally, the step of filtering target influencing factor data for the target category from the influencing factor data of each category, and adjusting the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data, to obtain the user traffic prediction network corresponding to each influencing factor data, includes:
[0013] Among the influencing factor data of each category, a self-attention network is used to select the target influencing factor data of the target category; the target category is the category of influencing factor data that has the greatest impact on user traffic during the target time period.
[0014] By adjusting the prediction parameters of the initial user traffic prediction network for each of the influencing factor data other than the target influencing factor data using the target influencing factor data, a user traffic prediction network corresponding to each of the influencing factor data is obtained.
[0015] Optionally, the step of selecting target influencing factor data for a target category from the influencing factor data of each category using a self-attention network includes:
[0016] For each category of influencing factor data, a self-attention network is used to calculate the historical impact evaluation value of the influencing factor data for that category over a historical period.
[0017] For a category with first influencing factor data for a target time period, the first influencing factor data for that category during the target time period is obtained. Based on the first influencing factor data, the historical impact evaluation value of the influencing factor data for that category is adjusted through the self-attention network to obtain the target impact evaluation value of the influencing factor data for that category.
[0018] For categories where no data for the first influencing factor exists for the target period, the historical impact evaluation value of the influencing factor data for that category is used as the target impact evaluation value of the influencing factor data for that category.
[0019] Among the target impact evaluation values of the impact factor data for each category, the impact factor data of the category with the largest target impact evaluation value is selected as the target impact factor data for the target category.
[0020] Optionally, for each category of influencing factor data, predicting user traffic data for a target time period based on the user traffic prediction network corresponding to the influencing factor data of that category includes:
[0021] When the influencing factor data of the aforementioned category is periodic data, the influencing factor data of the aforementioned category is divided into multiple groups of influencing factor data according to the time period, and each group of influencing factor data is input into the user traffic prediction network of the aforementioned category to obtain multiple initial user traffic data; the initial user traffic data are weighted and summed according to the weight of the time period to obtain the user traffic data of the influencing factor data of the aforementioned category in the target time period.
[0022] When the influencing factor data of the category is non-periodic, the user traffic data for the target time period corresponding to the influencing factor data of the category is predicted according to the user traffic prediction network of the category.
[0023] Optionally, determining the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factors includes:
[0024] The influencing factor data for each category is input into a fully connected neural network to obtain the weight ratio of the influencing factor data for each category.
[0025] The user traffic data for the target time period corresponding to the influencing factor data of each category is weighted and summed according to the weight ratio of the influencing factor data of each category to obtain the target user traffic data for the target time period.
[0026] Secondly, this application also provides a user traffic prediction device. The device includes:
[0027] The acquisition module is used to acquire data on multiple categories of influencing factors; the influencing factor data is the data corresponding to the influencing factors of user traffic within a historical time period.
[0028] The selection module is used to filter the target influencing factor data of the target category from the influencing factor data of each category, and adjust the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data, so as to obtain the user traffic prediction network corresponding to each influencing factor data.
[0029] The prediction module is used to predict the user traffic data for the target time period corresponding to the influencing factor data of each category, based on the user traffic prediction network corresponding to the influencing factor data of that category.
[0030] The determination module is used to determine the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factors.
[0031] Optionally, the device further includes:
[0032] The network acquisition module is used to acquire the data type of each type of influencing factor data, and select the user traffic prediction network that matches the data type of each type of influencing factor data from the various user traffic prediction networks, and use it as the user traffic prediction network corresponding to each type of influencing factor data.
[0033] Optionally, the selection module is specifically used for:
[0034] Among the influencing factor data of each category, a self-attention network is used to select the target influencing factor data of the target category; the target category is the category of influencing factor data that has the greatest impact on user traffic during the target time period.
[0035] By adjusting the prediction parameters of the initial user traffic prediction network for each of the influencing factor data other than the target influencing factor data using the target influencing factor data, a user traffic prediction network corresponding to each of the influencing factor data is obtained.
[0036] Optionally, the selection module is specifically used for:
[0037] For each category of influencing factor data, a self-attention network is used to calculate the historical impact evaluation value of the influencing factor data for that category over a historical period.
[0038] For a category with first influencing factor data for a target time period, the first influencing factor data for that category during the target time period is obtained. Based on the first influencing factor data, the historical impact evaluation value of the influencing factor data for that category is adjusted through the self-attention network to obtain the target impact evaluation value of the influencing factor data for that category.
[0039] For categories where no data for the first influencing factor exists for the target period, the historical impact evaluation value of the influencing factor data for that category is used as the target impact evaluation value of the influencing factor data for that category.
[0040] Among the target impact evaluation values of the impact factor data for each category, the impact factor data of the category with the largest target impact evaluation value is selected as the target impact factor data for the target category.
[0041] Optionally, the prediction module is specifically used for:
[0042] When the influencing factor data of the aforementioned category is periodic data, the influencing factor data of the aforementioned category is divided into multiple groups of influencing factor data according to the time period, and each group of influencing factor data is input into the user traffic prediction network of the aforementioned category to obtain multiple initial user traffic data; the initial user traffic data are weighted and summed according to the weight of the time period to obtain the user traffic data of the influencing factor data of the aforementioned category in the target time period.
[0043] When the influencing factor data of the category is non-periodic, the user traffic data for the target time period corresponding to the influencing factor data of the category is predicted according to the user traffic prediction network of the category.
[0044] Optionally, the determining module is specifically used for:
[0045] The influencing factor data for each category is input into a fully connected neural network to obtain the weight ratio of the influencing factor data for each category.
[0046] The user traffic data for the target time period corresponding to the influencing factor data of each category is weighted and summed according to the weight ratio of the influencing factor data of each category to obtain the target user traffic data for the target time period.
[0047] Thirdly, this application provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described in any one of the first aspects.
[0048] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in any one of the first aspects.
[0049] Fifthly, this application provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in any one of the first aspects.
[0050] The aforementioned user traffic prediction method, apparatus, computer equipment, and storage medium acquire multiple categories of influencing factor data; the influencing factor data refers to data corresponding to influencing factors of user traffic within historical time periods; from each category of influencing factor data, target category influencing factor data is filtered, and the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data is adjusted to obtain the user traffic prediction network corresponding to each category of influencing factor data; for each category of influencing factor data, the user traffic data for the target time period corresponding to the category of influencing factor data is predicted according to the user traffic prediction network corresponding to the category of influencing factor data; based on the user traffic prediction network corresponding to the target time period corresponding to each category of influencing factor data, the target user traffic data for the target time period is determined. By predicting the user traffic data for the target time period corresponding to multiple categories of influencing factor data, and determining the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each category of influencing factor data, the prediction of target user traffic avoids the situation where the prediction of target user traffic relies too much on a single influencing factor data. By comprehensively considering multiple influencing factor data, the prediction results obtained improve the accuracy of the predicted user traffic data. Attached Figure Description
[0051] Figure 1 This is a flowchart illustrating a user traffic prediction method in one embodiment;
[0052] Figure 2 This is a flowchart illustrating the adjustment steps of a user traffic prediction network in one embodiment.
[0053] Figure 3 This is a flowchart illustrating the steps for filtering target influencing factor data in one embodiment;
[0054] Figure 4 This is a flowchart illustrating an example of user traffic prediction in another embodiment;
[0055] Figure 5 This is a structural block diagram of a user traffic prediction device in one embodiment;
[0056] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0058] The user traffic prediction method provided in this application can be applied to terminals, servers, and systems including both terminals and servers, and is implemented through interaction between the terminal and the server. The terminal can include, but is not limited to, various personal computers, laptops, tablets, etc. This terminal predicts user traffic data for target time periods corresponding to multiple categories of influencing factor data, and determines the target user traffic data for the target time period based on the user traffic data for each of the influencing factor data for the target time period. This avoids the situation where the prediction of target user traffic relies too heavily on a single influencing factor data. By comprehensively considering multiple influencing factor data, the prediction result obtained improves the accuracy of the predicted user traffic data.
[0059] In one embodiment, such as Figure 1 As shown, a user traffic prediction method is provided. Taking the application of this method to a terminal as an example, the method includes the following steps:
[0060] Step S101: Obtain data on influencing factors for multiple categories.
[0061] Among them, the influencing factor data refers to the data corresponding to the factors affecting user traffic within a historical period.
[0062] In this embodiment, with authorization from the branch, the terminal retrieves data corresponding to multiple categories of factors influencing user traffic from the branch's historical database for historical periods. These influencing factors may include, but are not limited to, historical traffic data (average foot traffic last week / month / quarter / year, number of visitors in the same period last year, etc.); date and weather (date, temperature, humidity, wind force, air quality, etc.); branch location (location's accessibility, surrounding population); branch promotion policies (whether it's the beginning or end of the quarter, whether there are product promotions, promotion intensity, and promotion cycle, etc.); and branch online popularity (number of online visitors, number of online appointments, number of online customer service representatives, etc.).
[0063] Step S102: In the influencing factor data of each category, filter the target influencing factor data of the target category, and adjust the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data to obtain the user traffic prediction network corresponding to each influencing factor data.
[0064] In this embodiment, the terminal randomly selects one category of influencing factor data from the acquired influencing factor data of each category as the target influencing factor data for the target category. The terminal adjusts the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data based on the target influencing factor data of the target analogy, thus obtaining the user traffic prediction network corresponding to each influencing factor data. The user traffic prediction network can be, but is not limited to, any neural network capable of predicting data trends for a target time period based on historical data, such as Long Short-Term Memory (LSTM) networks, Deep Neural Networks (DNNs), etc. The specific adjustment process will be explained in detail later.
[0065] Step S103: For each category of influencing factor data, predict the user traffic data for the target time period corresponding to the category of influencing factor data based on the user traffic prediction network corresponding to the category of influencing factor data.
[0066] In this embodiment, the terminal inputs the influencing factor data for each category into the user traffic prediction network corresponding to that category, thereby predicting the user traffic data for the target time period under the influence of that category of influencing factors. The specific prediction process of each user traffic prediction network will be described in detail later.
[0067] Step S104: Determine the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to the data of each influencing factor.
[0068] In this embodiment, the terminal performs weighted averaging on the user traffic data for the target time period corresponding to each influencing factor through a fully connected network to obtain the target shortest target user traffic data.
[0069] Based on the above scheme, by predicting the user traffic data for the target time period corresponding to various categories of influencing factor data, and determining the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factor data, the prediction of target user traffic is avoided from relying too much on a single influencing factor data. By comprehensively considering multiple influencing factor data, the prediction results obtained improve the accuracy of the predicted user traffic data.
[0070] Optionally, after obtaining data on influencing factors across multiple categories, the following may also be included:
[0071] Obtain the data type of each type of influencing factor data, and select the user traffic prediction network that matches the data type of each type of influencing factor data from among the user traffic prediction networks, and use it as the user traffic prediction network corresponding to each type of influencing factor data.
[0072] In this embodiment, after acquiring multiple categories of influencing factor data, the terminal obtains the data type of each category of influencing factor data through various influencing factor databases. For each category of influencing factor data data, the terminal filters out the user traffic prediction network corresponding to that data type from among the various user traffic prediction networks, thus obtaining the user traffic prediction network corresponding to each category of influencing factor data. The data type refers to the category of data characteristics of the influencing factor data, including time series, increasing / decreasing, and fluctuation types. For example, if the category of influencing factor data is historical traffic data, the corresponding data type for that category is time series.
[0073] Specifically, when the influencing factor is historical traffic data, since the data type of historical traffic data is time series, the terminal uses an LSTM network to make predictions for this type of data. When the influencing factors include date and weather, branch location, branch promotion policies, and online popularity of the branch, the terminal can use a DNN network to make predictions.
[0074] Based on the above scheme, according to the data type of each influencing factor, the user traffic prediction network corresponding to each type of influencing factor data is selected, which improves the accuracy of predicting short-term user traffic based on each type of influencing factor data.
[0075] Optional, such as Figure 2 As shown, among the influencing factor data of each category, target influencing factor data for the target category is filtered out, and the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data is adjusted to obtain the user traffic prediction network corresponding to each influencing factor data, including:
[0076] Step S201: From the influencing factor data of each category, select the target influencing factor data of the target category through a self-attention network.
[0077] Among them, the target category is the category of data that has the greatest impact on user traffic during the target time period.
[0078] In this embodiment, the terminal uses a self-attention network to calculate the attention score of the influencing factor data for each category, and selects the influencing factor data of the category with the highest attention score as the target influencing factor data for the target analogy. The self-attention network can be any neural network based on attention (self-attention mechanism) capable of implementing the above steps.
[0079] Specifically, the formula for the above self-attention mechanism network is shown below:
[0080] Attention scoring formula:
[0081] In the above formula, the number of queries is: Key: XW e This relates to the input of the attention model, where X is the data output by each module through the neural network, i.e., X = 1stm(input) or X = dnn(input), and the value is XW. v In the parameters above, the query volume represents the input influencing factor data, the key value represents the category of the influencing factor data, and the value is the parameter representing the importance of the influencing factor data.
[0082] Step S202: Using the target influencing factor data, adjust the prediction parameters of the initial user traffic prediction network for each influencing factor data other than the target influencing factor data to obtain the user traffic prediction network corresponding to each influencing factor data.
[0083] In this embodiment, the terminal adjusts the prediction parameters of the user traffic prediction network corresponding to the influencing factor data of each non-target influencing factor data using the target influencing factor data, thereby obtaining the user traffic prediction network corresponding to each influencing factor data. The terminal does not adjust the initial user traffic prediction model corresponding to the target influencing factor data, and directly uses the initial user traffic prediction model corresponding to the target influencing factor data as the user traffic prediction network corresponding to that target influencing factor data.
[0084] Specifically, the activated RIM (i.e., the initial user traffic prediction network for target influencing factor data) exchanges information (i.e., target influencing factor data) with other RIM modules (i.e., the initial user traffic prediction network for influencing factor data of non-target influencing factor data), achieving sparse information transmission. Other RIM modules adjust their prediction parameters based on the exchanged information. The state of the k-th RIM module at time t+1 is as follows:
[0085]
[0086]
[0087]
[0088]
[0089] In the above formula, The prediction parameter values of the initial user traffic prediction network for different influencing factors, h_(t+1,k), h ′ t,k The data represent the target influencing factors at time h'(t) and time h(t+1).
[0090] Based on the above scheme, the terminal improves the prediction accuracy of the user traffic prediction network by selecting target influencing factor data and adjusting the influencing factor data of each non-target influencing factor data.
[0091] Optional, such as Figure 3 As shown, among the influencing factor data of each category, the target influencing factor data for the target category is selected through a self-attention network, including:
[0092] Step S301: For each category of influencing factor data, calculate the historical impact evaluation value of the category's influencing factor data in the historical time period using a self-attention network.
[0093] In this embodiment, the terminal calculates the historical impact evaluation value of each category of influencing factor data over a historical period using a self-attention network. The historical impact evaluation value is an assessment of the degree of influence of the category corresponding to the influencing factor data on user traffic over a historical period; a higher evaluation value indicates a more significant impact of the influencing factor data of that category on user traffic.
[0094] Step S302: For categories with first influencing factor data for the target time period, obtain the first influencing factor data for the category during the target time period, and adjust the historical impact evaluation value of the category's influencing factor data through a self-attention network based on the first influencing factor data to obtain the target impact evaluation value of the category's influencing factor data; for categories without first influencing factor data for the target time period, use the historical impact evaluation value of the category's influencing factor data as the target impact evaluation value of the category's influencing factor data.
[0095] In this embodiment, the terminal queries the database to see if there is a category of first influencing factor data for the target time period. The category of first influencing factor data is a preset category of influencing factor data in the terminal database. This first influencing factor data category is a special category that significantly impacts user traffic, such as unavoidable natural disasters (volcanic eruptions, earthquakes, torrential rains, typhoons, tsunamis, etc.) or activity periods (promotional seasons, participation-based prizes, buy-one-get-one-free promotions, etc.).
[0096] When a category has data on the first influencing factor for the target time period, the terminal acquires the data on the first influencing factor for that category during the target time period. Based on this data, it adjusts the historical impact evaluation values of the category's influencing factor data using a self-attention network to obtain the target impact evaluation values for all categories of influencing factor data. For categories where no data on the first influencing factor for the target time period exists, the historical impact evaluation values of each category's influencing factor data are used as the target impact evaluation values for each category's influencing factor data.
[0097] Step S303: Among the target impact evaluation values of the impact factor data of each category, select the impact factor data of the category with the largest target impact evaluation value as the target impact factor data of the target category.
[0098] In this embodiment, the terminal selects the category of influencing factor data with the largest target influence evaluation value from the target influence evaluation values of each category of influencing factor data, and uses it as the target influencing factor data of the target category.
[0099] Based on the above scheme, the terminal improves the prediction accuracy of each user traffic prediction network by selecting the data of the factors that have the greatest impact on user traffic as the target influencing factor data.
[0100] Optionally, for each category of influencing factor data, based on the user traffic prediction network corresponding to the category's influencing factor data, the user traffic data for the target time period corresponding to the category's influencing factor data is predicted. This includes: when the category's influencing factor data is periodic, dividing the category's influencing factor data into multiple groups of influencing factor data according to the time period, and inputting each group of influencing factor data into the category's user traffic prediction network to obtain multiple initial user traffic data; weighting and summing each initial user traffic data according to the weight of the time period to obtain the user traffic data for the category's influencing factor data in the target time period; when the category's influencing factor data is non-periodic, predicting the user traffic data for the target time period corresponding to the category's influencing factor data based on the category's user traffic prediction network.
[0101] In this embodiment, when the influencing factor data of this category is periodic data, the terminal divides the influencing factor data of this category into multiple groups of influencing factor data according to the time period, wherein each group of influencing factor data is influencing factor data within one time period. The terminal inputs each group of influencing factor data into the user traffic prediction network of this category to obtain multiple initial user traffic data. Among them, the periodic data is data of time series type. The terminal calculates the weight of the influencing factor data of each time period based on the influencing factor data of each time period and the number of groups of all influencing factor data. The terminal performs a weighted summation of each initial user traffic data according to the weight of the influencing factor data of each time period to obtain the user traffic data of this category of influencing factor data in the target time period.
[0102] When the influencing factor data for this category is non-periodic, the terminal inputs the influencing factor data for this category into the user traffic prediction network for that category to predict the user traffic data for the target time period corresponding to the influencing factor data for this category.
[0103] Based on the above scheme, by applying different prediction methods to different categories of influencing factor data, the prediction accuracy of each category of influencing factor data has been improved.
[0104] Optionally, the target user traffic data for the target time period is determined based on the user traffic data for the target time period corresponding to the data of each influencing factor, including: inputting the data of each category of influencing factors into a fully connected neural network to obtain the weight ratio of the data of each category of influencing factors; and performing weighted summation on the user traffic data for the target time period corresponding to the data of each category of influencing factors according to the weight ratio of the data of each category of influencing factors to obtain the target user traffic data for the target time period.
[0105] In this embodiment, before performing step S101, the terminal inputs the sample data of influencing factors for each category into the user prediction network corresponding to the influencing factor data for each category, thereby obtaining sample user traffic data for each category's sample time period. The terminal inputs the sample user traffic data for each category's sample time period, along with the actual user traffic data for each category's sample time period, into an initial fully connected neural network to train the weight calculation parameters of this fully connected neural network. Based on the influencing factor data for each category, the terminal calculates the weight ratio of the influencing factor data for each category using the trained fully connected neural network. Finally, the terminal performs a weighted summation of the user traffic data for the target time period corresponding to the influencing factor data for each category, according to the weight ratio of the influencing factor data for each category, to obtain the target user traffic data for the target time period.
[0106] Based on the above scheme, by predicting the user traffic data for the target time period corresponding to various categories of influencing factor data, and determining the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factor data, the accuracy of the prediction is improved.
[0107] This application also provides a user traffic prediction example, as shown in Figure 4. The specific processing steps include the following:
[0108] Step S401: Obtain data on influencing factors for multiple categories.
[0109] Step S402: Obtain the data type of each type of influencing factor data, and select the user traffic prediction network that matches the data type of each type of influencing factor data from the user traffic prediction networks, as the user traffic prediction network corresponding to each type of influencing factor data.
[0110] Step S403: For each category of influencing factor data, calculate the historical impact evaluation value of the influencing factor data of that category in the historical time period using a self-attention network.
[0111] Step S404: For a category with first influencing factor data for the target time period, obtain the first influencing factor data of the category during the target time period, and adjust the historical impact evaluation value of the influencing factor data of the category through the self-attention network based on the first influencing factor data to obtain the target impact evaluation value of the influencing factor data of the category.
[0112] Step S405: For categories where no data for the first influencing factor in the target time period exists, the historical impact evaluation value of the influencing factor data in that category is used as the target impact evaluation value of the influencing factor data in that category.
[0113] Step S406: Among the target impact evaluation values of the impact factor data of each category, select the impact factor data of the category with the largest target impact evaluation value as the target impact factor data of the target category.
[0114] Step S407: Using the target influencing factor data, adjust the prediction parameters of the initial user traffic prediction network for each influencing factor data other than the target influencing factor data to obtain the user traffic prediction network corresponding to each influencing factor data.
[0115] Step S408: For each category of influencing factor data, if the influencing factor data of the category is periodic data, divide the influencing factor data of the category into multiple groups of influencing factor data according to the time period, and input each group of influencing factor data into the user traffic prediction network of the category to obtain multiple initial user traffic data.
[0116] Step S409: The initial user traffic data are weighted and summed according to the weight of the time period to obtain the user traffic data of the influencing factor data of the category in the target time period.
[0117] Step S410: If the influencing factor data of the category is non-periodic data, predict the user traffic data for the target time period corresponding to the influencing factor data of the category according to the user traffic prediction network of the category.
[0118] Step S411: Input the influencing factor data of each category into a fully connected neural network to obtain the weight ratio of the influencing factor data of each category.
[0119] Step S412: The user traffic data for the target time period corresponding to the influencing factor data of each category is weighted and summed according to the weight ratio of the influencing factor data of each category to obtain the target user traffic data for the target time period.
[0120] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0121] Based on the same inventive concept, this application also provides a user traffic prediction device for implementing the user traffic prediction method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more user traffic prediction device embodiments provided below can be found in the limitations of the user traffic prediction method described above, and will not be repeated here.
[0122] In one embodiment, such as Figure 5 As shown, a user traffic prediction device is provided, including: an acquisition module 510, a selection module 520, a prediction module 530, and a determination module 540, wherein:
[0123] The acquisition module 510 is used to acquire data on multiple categories of influencing factors; the influencing factor data is the data corresponding to the influencing factors of user traffic in historical time periods.
[0124] The selection module 520 is used to filter the target influencing factor data of the target category from the influencing factor data of each category, and adjust the initial user traffic prediction network corresponding to the influencing factor data other than the target influencing factor data, so as to obtain the user traffic prediction network corresponding to each influencing factor data.
[0125] Prediction module 530 is used to predict user traffic data for a target time period corresponding to the influencing factor data of each category, based on the user traffic prediction network corresponding to the influencing factor data of the category.
[0126] The determination module 540 is used to determine the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factor data.
[0127] Optionally, the device further includes:
[0128] The network acquisition module is used to acquire the data type of each type of influencing factor data, and select the user traffic prediction network that matches the data type of each type of influencing factor data from the various user traffic prediction networks, and use it as the user traffic prediction network corresponding to each type of influencing factor data.
[0129] Optionally, the selection module 520 is specifically used for:
[0130] Among the influencing factor data of each category, a self-attention network is used to select the target influencing factor data of the target category; the target category is the category of influencing factor data that has the greatest impact on user traffic during the target time period.
[0131] By adjusting the prediction parameters of the initial user traffic prediction network for each of the influencing factor data other than the target influencing factor data using the target influencing factor data, a user traffic prediction network corresponding to each of the influencing factor data is obtained.
[0132] Optionally, the selection module 520 is specifically used for:
[0133] For each category of influencing factor data, a self-attention network is used to calculate the historical impact evaluation value of the influencing factor data for that category over a historical period.
[0134] For a category with first influencing factor data for a target time period, the first influencing factor data for that category during the target time period is obtained. Based on the first influencing factor data, the historical impact evaluation value of the influencing factor data for that category is adjusted through the self-attention network to obtain the target impact evaluation value of the influencing factor data for that category.
[0135] For categories where no data for the first influencing factor exists for the target period, the historical impact evaluation value of the influencing factor data for that category is used as the target impact evaluation value of the influencing factor data for that category.
[0136] Among the target impact evaluation values of the impact factor data for each category, the impact factor data of the category with the largest target impact evaluation value is selected as the target impact factor data for the target category.
[0137] Optionally, the prediction module 530 is specifically used for:
[0138] When the influencing factor data of the aforementioned category is periodic data, the influencing factor data of the aforementioned category is divided into multiple groups of influencing factor data according to the time period, and each group of influencing factor data is input into the user traffic prediction network of the aforementioned category to obtain multiple initial user traffic data; the initial user traffic data are weighted and summed according to the weight of the time period to obtain the user traffic data of the influencing factor data of the aforementioned category in the target time period.
[0139] When the influencing factor data of the category is non-periodic, the user traffic data for the target time period corresponding to the influencing factor data of the category is predicted according to the user traffic prediction network of the category.
[0140] Optionally, the determining module 540 is specifically used for:
[0141] The influencing factor data for each category is input into a fully connected neural network to obtain the weight ratio of the influencing factor data for each category.
[0142] The user traffic data for the target time period corresponding to the influencing factor data of each category is weighted and summed according to the weight ratio of the influencing factor data of each category to obtain the target user traffic data for the target time period. Each module in the above-mentioned user traffic prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0143] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a user traffic prediction method. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.
[0144] Those skilled in the art will understand that the structure shown in Figure Y is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or may combine certain components, or may have different component arrangements.
[0145] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any one of the first aspects.
[0146] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects.
[0147] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described in any one of the first aspects.
[0148] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.
[0149] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0150] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0151] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A user traffic prediction method, characterized in that, The method includes: Acquire data on multiple categories of influencing factors; the influencing factor data is data corresponding to the influencing factors of user traffic within a historical period; the categories of influencing factor data include historical traffic data, date and weather, branch location, branch promotion policy, and online popularity of branches; Among the influencing factor data of each category, a self-attention network is used to select the target influencing factor data of the target category; the target category is the category of influencing factor data that has the greatest impact on user traffic during the target time period. By adjusting the prediction parameters of the initial user traffic prediction network for each influencing factor data other than the target influencing factor data using the target influencing factor data, the user traffic prediction network corresponding to each influencing factor data is obtained. For each category of influencing factor data, based on the user traffic prediction network corresponding to the influencing factor data of that category, predict the user traffic data for the target time period corresponding to the influencing factor data of that category; Based on the user traffic data for the target time period corresponding to each of the aforementioned influencing factors, determine the target user traffic data for the target time period; The selection of target influencing factor data for a target category from the influencing factor data of each category, using a self-attention network, includes: For each category of influencing factor data, a self-attention network is used to calculate the historical impact evaluation value of the influencing factor data for that category over a historical period. For a category with first influencing factor data for a target time period, the first influencing factor data for that category during the target time period is obtained. Based on the first influencing factor data, the historical impact evaluation value of the influencing factor data for that category is adjusted through the self-attention network to obtain the target impact evaluation value of the influencing factor data for that category. For categories where no data for the first influencing factor exists for the target period, the historical impact evaluation value of the influencing factor data for that category is used as the target impact evaluation value of the influencing factor data for that category. Among the target impact evaluation values of the impact factor data for each category, the impact factor data of the category with the largest target impact evaluation value is selected as the target impact factor data for the target category.
2. The method according to claim 1, characterized in that, After obtaining data on influencing factors across multiple categories, the process also includes: Obtain the data type of each type of influencing factor data, and select the user traffic prediction network that matches the data type of each type of influencing factor data from among the user traffic prediction networks, and use it as the user traffic prediction network corresponding to each type of influencing factor data.
3. The method according to claim 1, characterized in that, For each category of influencing factor data, based on the user traffic prediction network corresponding to the influencing factor data of that category, the user traffic data for the target time period corresponding to the influencing factor data of that category is predicted, including: When the influencing factor data of the aforementioned category is periodic data, the influencing factor data of the aforementioned category is divided into multiple groups of influencing factor data according to the time period, and each group of influencing factor data is input into the user traffic prediction network of the aforementioned category to obtain multiple initial user traffic data; the initial user traffic data are weighted and summed according to the weight of the time period to obtain the user traffic data of the influencing factor data of the aforementioned category in the target time period. When the influencing factor data of the category is non-periodic, the user traffic data for the target time period corresponding to the influencing factor data of the category is predicted according to the user traffic prediction network of the category.
4. The method according to claim 1, characterized in that, The step of determining the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factors includes: The influencing factor data for each category is input into a fully connected neural network to obtain the weight ratio of the influencing factor data for each category. The user traffic data for the target time period corresponding to the influencing factor data of each category is weighted and summed according to the weight ratio of the influencing factor data of each category to obtain the target user traffic data for the target time period.
5. A user traffic prediction device, characterized in that, The device includes: The acquisition module is used to acquire data on multiple categories of influencing factors. The influencing factor data is data corresponding to the influencing factors of user traffic within a historical period. The categories of the influencing factor data include historical traffic data, date and weather, branch location, branch promotion policy, and online popularity of the branch. The selection module is used to select target influencing factor data of a target category from the influencing factor data of each category through a self-attention network; the target category is the category of influencing factor data that has the greatest impact on user traffic during the target time period; the prediction parameters of the initial user traffic prediction network of each influencing factor data other than the target influencing factor data are adjusted through the target influencing factor data to obtain the user traffic prediction network corresponding to each influencing factor data. The prediction module is used to predict the user traffic data for the target time period corresponding to the influencing factor data of each category, based on the user traffic prediction network corresponding to the influencing factor data of that category. The determination module is used to determine the target user traffic data for the target time period based on the user traffic data for the target time period corresponding to each of the influencing factors data; The selection module is specifically used to calculate the historical impact evaluation value of the influencing factor data of each category in a historical period using a self-attention network; for categories with first influencing factor data in a target period, obtain the first influencing factor data of the category in the target period, and adjust the historical impact evaluation value of the influencing factor data of the category using the self-attention network based on the first influencing factor data to obtain the target impact evaluation value of the influencing factor data of the category; for categories without first influencing factor data in a target period, use the historical impact evaluation value of the influencing factor data of the category as the target impact evaluation value of the influencing factor data of the category; and select the influencing factor data of the category with the largest target impact evaluation value from the target impact evaluation values of the influencing factor data of each category as the target influencing factor data of the target category.
6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.
8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.