CN-122028104-A - Method, apparatus, device, storage medium and program product for traffic prediction

CN122028104ACN 122028104 ACN122028104 ACN 122028104ACN-122028104-A

Abstract

The application discloses a flow prediction method, a flow prediction device, flow prediction equipment, a storage medium and a program product. The method comprises the steps of obtaining crowd transfer trend and historical flow data of base stations according to each base station, constructing a crowd transfer adjacent matrix according to the crowd transfer trend, wherein elements of the crowd transfer adjacent matrix represent the crowd to column base station transfer trend of a row base station, the row base station is a base station corresponding to a row where the elements are located, the column base station is a base station corresponding to a column where the elements are located, constructing a flow transfer matrix according to the historical flow data, calculating the sum value of the flow transfer matrix and the crowd transfer adjacent matrix to obtain a base station adjacent weight matrix, constructing a data matrix according to the historical flow data, and carrying out flow prediction according to the base station adjacent weight matrix and the data matrix to obtain a flow prediction result. Therefore, the two dimensions of the flow transfer trend and the crowd transfer trend are integrated, the flow transfer trend between the base stations can be accurately judged, and the accuracy of flow prediction is improved.

Inventors

CHEN XIAOMING

Assignees

中国移动通信集团湖南有限公司
中国移动通信集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260205

Claims (11)

1. A method of traffic prediction, for application to an electronic device, the method comprising: Aiming at each base station, acquiring crowd transfer trend and historical flow data of the base station, wherein the crowd transfer trend is predicted according to moving tracks of crowds; According to the crowd transfer trend, constructing a crowd transfer adjacent matrix, wherein elements of the crowd transfer adjacent matrix represent the crowd transfer trend of row base stations to column base stations, the row base stations are base stations corresponding to rows where the elements are located, and the column base stations are base stations corresponding to columns where the elements are located; Constructing a flow transfer matrix according to the historical flow data, wherein elements of the flow transfer matrix represent the flow of the row base station and the flow of the column base station transfer trend; calculating the sum of the traffic transfer matrix and the group transfer adjacency matrix to obtain a base station adjacency weight matrix; Constructing a data matrix according to the historical flow data, wherein elements in the data matrix are used for representing flow data of a column base station at a row historical moment, and the row historical moment is a historical moment corresponding to a row where the elements are located; And carrying out flow prediction according to the base station adjacent weight matrix and the data matrix to obtain a flow prediction result.
2. The method of claim 1, wherein the obtaining the population transfer trend of the base station comprises: acquiring historical user distribution data of each sub-period of the base station in a historical period; Clustering the historical user distribution data aiming at each historical user distribution data to obtain a cluster; determining the moving track of each clustering center according to the clustering center of the corresponding clustering cluster of the base station in each subinterval; and determining the crowd transfer trend of the base station according to the movement track.
3. The method of claim 2, wherein constructing a population transfer adjacency matrix based on the population transfer trend comprises: for each base station, determining a weight coefficient of the base station according to user distribution data in each cluster of the base station; and taking the weight coefficient as the value of an element in the group transfer adjacent matrix to obtain the group transfer adjacent matrix.
4. A method according to claim 3, wherein said determining the weight coefficient of the base station from the user distribution data in each cluster of the base station comprises: for each cluster, calculating the distance between each user distribution data in the cluster and the cluster center of the cluster; Summing the distances corresponding to the distributed data of each user aiming at each cluster to obtain a first sum value; and calculating an average value according to the first sum value corresponding to each cluster to obtain the weight coefficient.
5. The method of claim 1, wherein the performing traffic prediction according to the base station adjacency weight matrix and the data matrix to obtain a traffic prediction result comprises: Inputting the data matrix into a feature extraction network in a flow prediction model, and extracting features of the data matrix through the feature extraction network according to the base station adjacent weight matrix to obtain coding features; Inputting the coding features into a multi-layer feedforward neural network in the flow prediction model, calculating the mean value and variance of the coding features through the multi-layer feedforward neural network, and calculating potential variables according to the mean value and variance; inputting the potential variables into a reconstruction network in the flow prediction model, and constructing reconstruction features according to the reconstruction network; and inputting the reconstructed features into a hidden layer in the flow prediction model, and carrying out weighted summation on the features with different dimensions in the reconstructed features through the hidden layer to obtain the flow prediction result.
6. The method of claim 5, wherein prior to said inputting the data matrix into the feature extraction network in the traffic prediction model, the method further comprises: Acquiring training data; Inputting the training data into a neural network model, and acquiring sample reconstruction characteristics output by a reconstruction network in the neural network model, sample flow prediction results output by the neural network model and sample coding characteristics output by a characteristic extraction network in the neural network model; calculating reconstruction loss according to the sample reconstruction characteristics and the sample coding characteristics; Calculating to obtain a mean square error loss and a relative entropy loss according to the sample flow prediction result and the label data of the training data; Carrying out weighted summation on the reconstruction loss, the mean square error loss and the relative entropy loss to obtain a total loss function; And adjusting the neural network model by using the total loss function to obtain the flow prediction model.
7. The method of claim 1, wherein after performing traffic prediction according to the base station adjacency weight matrix and the data matrix to obtain a traffic prediction result, the method further comprises: for each base station, acquiring actual flow data, wherein the actual flow data comprises the number of users using preset services, the total number of users and the actual total flow in the signal coverage area of the base station; calculating a correction coefficient according to the number of users, the total number of users, the actual total flow and the flow prediction result; Acquiring a future population transfer adjacency matrix at the next moment adjacent to the flow prediction result; and calculating the product of the correction coefficient and the future population transfer adjacency matrix to obtain a correction population transfer adjacency matrix, wherein the correction population transfer adjacency matrix is used for carrying out flow prediction at the next moment.
8. An apparatus for traffic prediction, for application to an electronic device, the apparatus comprising: The acquisition module is used for acquiring crowd transfer trend and historical flow data of each base station according to each base station, wherein the crowd transfer trend is predicted according to the moving track of the crowd; The construction module is used for constructing a group transfer adjacent matrix according to the group transfer trend, wherein elements of the group transfer adjacent matrix represent the group transfer trend of row base stations to column base stations, the row base stations are base stations corresponding to rows where the elements are located, and the column base stations are base stations corresponding to columns where the elements are located; the construction module is used for constructing a flow transfer matrix according to the historical flow data, and elements of the flow transfer matrix represent the flow transfer trend of the row base station to the column base station; The calculation module is used for calculating the sum value of the flow transfer matrix and the group transfer adjacency matrix to obtain a base station adjacency weight matrix; the construction module is used for constructing a data matrix according to the historical flow data, wherein elements in the data matrix are used for representing flow data of a column base station at a row historical moment, and the row historical moment is a historical moment corresponding to a row where the elements are located; And the prediction module is used for carrying out flow prediction according to the base station adjacent weight matrix and the data matrix to obtain a flow prediction result.
9. An electronic device comprising a processor and a memory storing computer program instructions; a method of flow prediction as claimed in any one of claims 1 to 7 when executed by a processor.
10. A computer storage medium having stored thereon computer program instructions which when executed by a processor implement a method of flow prediction as claimed in any one of claims 1 to 7.
11. A computer program product, characterized in that instructions in the computer program product, when executed by a processor of an electronic device, cause the electronic device to perform the method of flow prediction according to any of claims 1-7.

Description

Method, apparatus, device, storage medium and program product for traffic prediction Technical Field The application belongs to the technical field of artificial intelligence, and particularly relates to a flow prediction method, a flow prediction device, flow prediction equipment, a storage medium and a program product. Background In a mobile communication network, by predicting the flow of a base station and then dynamically allocating network resources according to the prediction result, network congestion can be avoided and the service quality of the base station can be improved. At present, algorithms such as a time-space-variant diagram self-encoder (Spatio-Temporal Variational Graph Auto-Encoder, ST-VGAE), a diagram convolution network (Graph Convolutional Network, GCN), a diagram neural network (Graph Neural Network, GNN) and a time-space diagram convolution network (Spatio-Temporal Graph Convolutional Network, ST-GCN) are used for carrying out flow prediction by combining preset adjacent matrixes. Wherein the adjacency matrix is used to represent the correlation between neighboring cellular network areas. However, in practical implementation, the flow direction of the crowd is changed in real time, and the flow load of different base stations is dynamically changed, so that the fixed adjacency matrix cannot accurately reflect the correlation between cellular network areas, and thus the accuracy of flow prediction is reduced. Disclosure of Invention The embodiment of the application provides a method, a device, equipment, a storage medium and a program product for flow prediction, which can integrate two dimensions of a flow transfer trend and a crowd transfer trend and accurately judge the flow transfer trend among base stations. And finally, carrying out flow prediction according to the historical flow data and the base station adjacency weight matrix, thereby improving the accuracy of flow prediction. In a first aspect, an embodiment of the present application provides a method for traffic prediction, applied to an electronic device, where the method includes: Aiming at each base station, acquiring crowd transfer trend and historical flow data of the base station, wherein the crowd transfer trend is predicted according to moving tracks of crowds; According to the crowd transfer trend, constructing a crowd transfer adjacent matrix, wherein elements of the crowd transfer adjacent matrix represent the crowd transfer trend of row base stations to column base stations, the row base stations are base stations corresponding to rows where the elements are located, and the column base stations are base stations corresponding to columns where the elements are located; Constructing a flow transfer matrix according to the historical flow data, wherein elements of the flow transfer matrix represent the flow of the row base station and the flow of the column base station transfer trend; calculating the sum of the traffic transfer matrix and the group transfer adjacency matrix to obtain a base station adjacency weight matrix; Constructing a data matrix according to the historical flow data, wherein elements in the data matrix are used for representing flow data of a column base station at a row historical moment, and the row historical moment is a historical moment corresponding to a row where the elements are located; And carrying out flow prediction according to the base station adjacent weight matrix and the data matrix to obtain a flow prediction result. In one possible implementation manner, the obtaining the crowd transfer trend of the base station includes: acquiring historical user distribution data of each sub-period of the base station in a historical period; Clustering the historical user distribution data aiming at each historical user distribution data to obtain a cluster; determining the moving track of each cluster center according to the cluster center of the corresponding s cluster in each subinterval of the base station; and determining the crowd transfer trend of the base station according to the movement track. In one possible implementation manner, the constructing a population transfer adjacency matrix according to the population transfer trend includes: for each base station, determining a weight coefficient of the base station according to user distribution data in each cluster of the base station; and taking the weight coefficient as the value of an element in the group transfer adjacent matrix to obtain the group transfer adjacent matrix. In one possible implementation manner, the determining the weight coefficient of the base station according to the user distribution data in each cluster of the base station includes: for each cluster, calculating the distance between each user distribution data in the cluster and the cluster center of the cluster; Summing the distances corresponding to the distributed data of each user aiming at each cluster to obtain a first sum value; and calculati