Base station traffic prediction method, device and equipment and readable storage medium

Abstract

The invention provides a base station flow prediction method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining the space environment characteristics of a target base station according to a city knowledge graph, and enabling the city knowledge graph to represent the target relation between the base stations, the target relationship comprises a distance relationship between the base stations, a flow trend relationship between the base stations, the similarity of flow sequences between the base stations and the similarity of POI distribution between the base stations; according to the urban knowledge graph and the historical traffic sequence of the target base station, obtaining time characteristics of the historical traffic sequence of the target base station; and obtaining a traffic prediction value of the target base station according to the space environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station. The space environment characteristics of the target base station are obtained based on the target relationship between the base stations, and the traffic prediction value of the target base station is obtained based on the environment characteristics and the time characteristics of the historical traffic sequence, so that the traffic prediction precision of the base station is improved, and a large amount of manpower does not need to be consumed.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a base station traffic prediction method, device, equipment and readable storage medium. Background Art

[0002] Mobile Data Traffic Prediction refers to estimating the cumulative traffic value of a base station in the future (usually in units of X min). The typical application scenario is in an urban base station network. Based on the traffic prediction results of each base station in the network, the communication operator sets reasonable resource allocation, shuts down or puts the base station into sleep mode in areas with low traffic, and reasonably allocates the mobile traffic demand to the surrounding base stations in areas with high traffic. In existing actual production applications, the prediction results can only be preliminarily estimated based on the historical traffic values ​​of each base station, and then adjusted based on manual experience. Faced with a large number of base stations, it requires a lot of manpower and the prediction accuracy cannot be guaranteed. Summary of the invention

[0003] The purpose of the present invention is to provide a base station traffic prediction method, device, equipment and readable storage medium to solve the problem that the base station traffic prediction solution in the related art requires a lot of manpower and the prediction accuracy cannot be guaranteed.

[0004] To achieve the above object, an embodiment of the present invention provides a base station traffic prediction method, comprising:

[0005] According to the city knowledge graph, the spatial environment characteristics of the target base station are obtained. The city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the flow trend relationship between the base stations, the similarity of the flow sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations.

[0006] According to the city knowledge graph and the historical traffic sequence of the target base station, a time feature of the historical traffic sequence of the target base station is obtained;

[0007] A traffic prediction value of the target base station is obtained according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

[0008] Optionally, acquiring the spatial environment characteristics of the target base station according to the city knowledge graph includes:

[0009] The spatial environment characteristics of the target base station are obtained based on the city knowledge graph, the knowledge graph representation learning model and the multi-relationship graph convolutional neural network.

[0010] Optionally, acquiring the spatial environment features of the target base station according to the city knowledge graph, the knowledge graph representation learning network, and the multi-relation graph convolutional neural network includes:

[0011] Based on the knowledge graph representation learning model, high-order singular value decomposition is performed on the fact triples in the city knowledge graph to obtain an embedding vector corresponding to the base station in the city knowledge graph, where the base stations in the city knowledge graph include the target base station;

[0012] The embedded vector is processed based on a multi-relational graph convolutional neural network RGCN to obtain the spatial environment characteristics of the target base station.

[0013] Optionally, acquiring the time characteristics of the historical traffic sequence of the target base station according to the city knowledge graph and the historical traffic sequence of the target base station includes:

[0014] According to the city knowledge graph, at least one similar base station of the target base station is obtained, the distance between the similar base station and the target base station is less than a preset distance, the traffic trends of the similar base station and the target base station correspond to the same type of traffic trends, and the similarity of the traffic sequence between the similar base station and the target base station is greater than a first preset similarity value, and the similarity of the POI distribution between the similar base station and the target base station is greater than a second preset similarity value;

[0015] According to the historical traffic sequence of the similar base stations and the historical traffic sequence of the target base station, a time feature of the historical traffic sequence of the target base station is acquired.

[0016] Optionally, acquiring the time feature of the historical traffic sequence of the target base station according to the historical traffic sequence of the similar base station and the historical traffic sequence of the target base station includes:

[0017] Performing one-dimensional causal convolution processing on the historical traffic sequences of the target base station and each of the similar base stations to obtain convolution time features of the historical traffic sequences of the target base station and each of the similar base stations;

[0018] Inputting the convolution time feature into a multi-relation graph convolutional neural network RGCN to obtain the aggregated time feature of the target base station and each of the similar base stations;

[0019] According to the attention mechanism in the Transformer and the aggregated time features, the time features of the historical traffic sequence of the target base station are obtained.

[0020] Optionally, obtaining the traffic prediction value of the target base station according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station includes:

[0021] According to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, the traffic prediction value of the target base station is obtained by utilizing the attention fusion mechanism between base stations and a multi-layer perceptron.

[0022] The embodiment of the present invention further provides a base station traffic prediction device, comprising:

[0023] The first acquisition module is used to acquire the spatial environment characteristics of the target base station according to the city knowledge graph, wherein the city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the flow trend relationship between the base stations, the similarity of the flow sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations;

[0024] A second acquisition module is used to acquire the time characteristics of the historical traffic sequence of the target base station according to the city knowledge graph and the historical traffic sequence of the target base station;

[0025] The third acquisition module is used to obtain the traffic prediction value of the target base station according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

[0026] The embodiment of the present invention also provides a base station traffic prediction device, including: a transceiver and a processor;

[0027] The processor is used to obtain spatial environmental characteristics of the target base station based on the city knowledge graph, the city knowledge graph represents the target relationship between base stations, the target relationship includes the distance relationship between base stations, the traffic trend relationship between base stations, the similarity of traffic sequences between base stations, and the similarity of point of interest (POI) distribution between base stations; based on the city knowledge graph and the historical traffic sequence of the target base station, obtain the time characteristics of the historical traffic sequence of the target base station; based on the spatial environmental characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, obtain the traffic prediction value of the target base station.

[0028] An embodiment of the present invention also provides a base station traffic prediction device, comprising: a transceiver, a processor, a memory, and a program or instruction stored in the memory and executable on the processor; characterized in that when the processor executes the program or instruction, the steps of the base station traffic prediction method as described above are implemented.

[0029] An embodiment of the present invention further provides a readable storage medium on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the base station traffic prediction method as described above are implemented.

[0030] The beneficial effects of the above technical solution of the present invention are as follows:

[0031] In an embodiment of the present invention, the spatial environment characteristics of the target base station are obtained based on the city knowledge graph, and the city knowledge graph represents the target relationship between base stations, and the target relationship includes the distance relationship between base stations, the traffic trend relationship between base stations, the similarity of traffic sequences between base stations, and the similarity of the distribution of points of interest (POI) between base stations; based on the city knowledge graph and the historical traffic sequence of the target base station, the time characteristics of the historical traffic sequence of the target base station are obtained; based on the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, the traffic prediction value of the target base station is obtained. Through the above scheme, the spatial environment characteristics of the target base station are obtained based on the target relationship between base stations, and the traffic prediction value of the target base station is obtained based on the environmental characteristics and the time characteristics of the historical traffic sequence, which improves the accuracy of base station traffic prediction without consuming a lot of manpower. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is one of the flow charts of the base station traffic prediction method according to an embodiment of the present invention;

[0033] Figure 2 It is a processing flow chart of the environmental information modeling part in an embodiment of the present invention;

[0034] Figure 3 It is a processing flow chart of the time series modeling part in an embodiment of the present invention;

[0035] Figure 4 This is the second flowchart of the base station traffic prediction method according to an embodiment of the present invention;

[0036] Figure 5 This is an overall flow chart of a base station traffic prediction method according to an embodiment of the present invention;

[0037] Figure 6 This is a module diagram of a base station traffic prediction device in an embodiment of the present invention;

[0038] Figure 7 This is one of the structural block diagrams of the base station traffic prediction device in an embodiment of the present invention;

[0039] Figure 8 This is the second structural block diagram of the base station traffic prediction device in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0041] It should be understood that the references to "one embodiment" or "an embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present invention. Therefore, the references to "in one embodiment" or "in an embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0042] In various embodiments of the present invention, it should be understood that the size of the serial numbers of the following processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0043] Additionally, the terms "system" and "network" are often used interchangeably herein.

[0044] In the embodiments provided in the present application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0045] In the related art, when using machine learning algorithms to complete the task of base station traffic prediction, it is mainly analyzed as a time series prediction problem, and the historical traffic value sequence of the base station is used to predict the traffic in the future. Commonly used traditional methods include Support Vector Regression Model (SVR model for short), Auto-Regression Integrated Moving Average Model (ARIMA model for short), etc. These methods stabilize the non-stationary time series data and calculate the relevant parameters of the sequence for subsequent regression analysis. However, this type of method only predicts the next sequence based on the past time series, and does not take into account the impact of the spatial characteristics of the base station on the traffic. Therefore, the prediction accuracy is poor, and the similarity between base stations is not utilized, resulting in the trained prediction model only for one sequence and cannot be generalized.

[0046] The goal of the present invention is to establish a traffic prediction system based on the base station level. The base station is a public mobile communication base station, which is an interface device for mobile devices to access the Internet. Considering the spatial characteristics of the base station and the temporal characteristics of the traffic, the neural network model is implemented to greatly improve the accuracy of the future traffic prediction of the base station. The system is divided into two parts, which respectively mine the spatial characteristics and temporal characteristics of the base station. On this basis, the attention mechanism is used to make the neural network pay more attention to more effective features, and then predict the future traffic to achieve better prediction performance.

[0047] like Figure 1 As shown, an embodiment of the present invention provides a base station traffic prediction method, including:

[0048] Step 101: According to the city knowledge graph, the spatial environment characteristics of the target base station are obtained. The city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the traffic trend relationship between the base stations, the similarity of the traffic sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations.

[0049] In the embodiment of the present invention, the distance relationship indicates that two base stations are located within a certain distance, and the closer the distance, the more frequent the traffic interaction between them. If a user cuts off the connection with a base station, it is more likely to connect to a nearby base station.

[0050] For the above-mentioned traffic trend relationship, each base station has its own traffic trend, and this traffic trend will not change over time, and its traffic value will fluctuate up and down based on this trend. At the same time, this traffic trend is mainly affected by the location of the base station. In an embodiment of the present invention, the traffic trend of each base station is first normalized, and the traffic trends of the base stations are classified using a hierarchical clustering method. At the same time, in order to reduce the number of such relationships, it is further required that the traffic trends of the base stations are not only in the same category, but they also have the above-mentioned distance relationship. That is, the traffic trend relationship represents the traffic trend relationship between base stations located within a certain distance.

[0051] Regarding the similarity of traffic sequences between base stations, even if the traffic trends of two base stations are similar, their traffic sequences may still be very different, such as their absolute values ​​and the rate of traffic change. In order to further characterize the similarity of traffic sequences, the present invention uses the original traffic sequences of the base stations and dynamic time warping (DTW) to calculate the similarity between the traffic sequences of two base stations, and for each base station, selects the most similar k ones and considers that their traffic sequences are similar.

[0052] For the similarity of POI distribution between base stations, this similarity means that the POI distribution near the base station is similar. If the POI distribution between base stations is similar, then the traffic of the base stations is more likely to be similar. First, the number of POIs of each type around each base station is calculated, and then the cosine similarity of its POI distribution vector is calculated. For each base station, the m POI distributions that are most similar to it are selected, and the POI distributions between them are considered similar.

[0053] In the embodiment of the present invention, the spatial environment characteristics of the target base station are obtained through the environment information modeling part, and the environment information modeling part is composed of TuckER and a multi-relational graph convolutional neural network.

[0054] Step 102: According to the city knowledge graph and the historical traffic sequence of the target base station, the time characteristics of the historical traffic sequence of the target base station are obtained.

[0055] In the embodiment of the present invention, the time characteristics of the historical traffic sequence of the target base station are obtained through the time series modeling part, and the time series modeling part includes a convolutional neural network, a machine learning model transformer and a multi-relational graph convolutional neural network.

[0056] Step 103: Obtain a traffic prediction value of the target base station according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

[0057] In an embodiment of the present invention, the spatial environment characteristics of the target base station are obtained based on the city knowledge graph, and the city knowledge graph represents the target relationship between base stations, and the target relationship includes the distance relationship between base stations, the traffic trend relationship between base stations, the similarity of traffic sequences between base stations, and the similarity of the distribution of points of interest (POI) between base stations; based on the city knowledge graph and the historical traffic sequence of the target base station, the time characteristics of the historical traffic sequence of the target base station are obtained; based on the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, the traffic prediction value of the target base station is obtained. Through the above scheme, the spatial environment characteristics of the target base station are obtained based on the target relationship between base stations, and the traffic prediction value of the target base station is obtained based on the environmental characteristics and the time characteristics of the historical traffic sequence, which improves the accuracy of base station traffic prediction without consuming a lot of manpower.

[0058] Optionally, acquiring the spatial environment characteristics of the target base station according to the city knowledge graph includes:

[0059] The spatial environment characteristics of the target base station are obtained based on the city knowledge graph, the knowledge graph representation learning model and the multi-relational graph convolutional neural network.

[0060] In an embodiment of the present invention, the knowledge graph representation learning model may be specifically a TuckER model, and the multi-relational graph convolutional neural network may be specifically a Relational Graph Convolutional Network (RGCN).

[0061] Optionally, acquiring the spatial environment features of the target base station according to the city knowledge graph, the knowledge graph representation learning network, and the multi-relation graph convolutional neural network includes:

[0062] Based on the knowledge graph representation learning model, high-order singular value decomposition is performed on the fact triples in the city knowledge graph to obtain an embedding vector corresponding to the base station in the city knowledge graph, where the base stations in the city knowledge graph include the target base station;

[0063] The embedded vector is processed based on a multi-relational graph convolutional neural network RGCN to obtain the spatial environment characteristics of the target base station.

[0064] In the embodiment of the present invention, the processing flow of the environment information modeling part is as follows: Figure 2 shown.

[0065] In the embodiment of the present invention, the knowledge graph representation learning model mines the spatial environment features in the urban knowledge graph to obtain the above-mentioned low-dimensional embedding vector. Specifically, each directed representation in the urban knowledge graph (i.e., the above-mentioned fact triple) is as follows: h ,r,e t}, where e h Indicates the head entity, e t represents the tail entity, and r represents the relationship between the head entity and the tail entity. The TuckER model is used to pre-train the low-dimensional vectors of base stations in the city knowledge graph, i.e., embedding.

[0066]

[0067] Among them, W is the core tensor in TuckER decomposition, represents the embedding vector.

[0068] Then, the multi-relation graph convolutional neural network RGCN is used to optimize and update the embedding of each entity to make it more suitable for the base station traffic prediction task. The calculation process of RGCN can be expressed as:

[0069]

[0070] in, represents the neighbor set of node i under relation r, c i,r is the normalization constant, l is the l-th layer RGCN, as well as represents the learnable parameters in the neural network, The central node embedding representation representing the upper layer output, The embedding representation of the neighboring nodes of the central node outputted by the upper layer is used, and the embedding representation of each entity obtained by pre-training is used as the input of the first layer, and the output of the last layer is used as the spatial environment feature of the base station. The calculation process of the above RGCN can be the same as that of the existing solution.

[0071] In the embodiment of the present invention, by optimizing and updating the embedding of each entity in RGCN, the difference between the predicted value and the true value can be reduced, making it more suitable for the base station traffic prediction task.

[0072] Optionally, acquiring the time characteristics of the historical traffic sequence of the target base station according to the city knowledge graph and the historical traffic sequence of the target base station includes:

[0073] According to the city knowledge graph, at least one similar base station of the target base station is obtained, the distance between the similar base station and the target base station is less than a preset distance, the traffic trends of the similar base station and the target base station correspond to the same type of traffic trends, and the similarity of the traffic sequence between the similar base station and the target base station is greater than a first preset similarity value, and the similarity of the POI distribution between the similar base station and the target base station is greater than a second preset similarity value;

[0074] According to the historical traffic sequence of the similar base stations and the historical traffic sequence of the target base station, a time feature of the historical traffic sequence of the target base station is acquired.

[0075] In an embodiment of the present invention, a base station map is extracted based on a city knowledge graph, and the base station map includes at least one similar base station to the target base station. Since the distance between the similar base station and the target base station is less than a preset distance, and the traffic trends of the similar base station and the target base station correspond to the same type of traffic trends, and the similarity of the traffic sequence between the similar base station and the target base station is greater than a first preset similarity value, and the similarity of the POI distribution between the similar base station and the target base station is greater than a second preset similarity value, based on the historical traffic sequence of the similar base station and the historical traffic sequence of the target base station, the accuracy of the time characteristics of the historical traffic sequence of the target base station can be more accurately obtained, and then based on the time characteristics of the historical traffic sequence of the target base station and the spatial environment characteristics of the target base station, the accuracy of the future traffic estimation of the base station can be greatly improved.

[0076] Optionally, acquiring the time feature of the historical traffic sequence of the target base station according to the historical traffic sequence of the similar base station and the historical traffic sequence of the target base station includes:

[0077] Performing one-dimensional causal convolution processing on the historical traffic sequences of the target base station and each of the similar base stations to obtain convolution time features of the historical traffic sequences of the target base station and each of the similar base stations;

[0078] Inputting the convolution time feature into a multi-relation graph convolutional neural network RGCN to obtain the aggregated time feature of the target base station and each of the similar base stations;

[0079] According to the attention mechanism in the Transformer and the aggregated time features, the time features of the historical traffic sequence of the target base station are obtained.

[0080] In the embodiment of the present invention, when obtaining the time characteristics of the historical traffic sequence of the base station, time series modeling is used, and the time series modeling uses CNN, transformer and RGCN. In the time series modeling part, Figure 3 As shown in the figure, we first use one-dimensional causal convolution to capture the temporal features in the base station's historical traffic sequence. Causality can ensure that the temporal features of the current moment are only determined by the current moment and the moments before it, and have nothing to do with future moments. At the same time, the temporal features obtained by the convolution operation are the current local shape features, which helps to focus on the shape features rather than the features of a single point in the subsequent attention mechanism. The calculation process of one-dimensional causal convolution can be expressed as:

[0081]

[0082] Among them, k represents the size of the convolution kernel, and b l represents the learnable parameters of the neural network, It represents the time characteristics of the output of the upper (l-1-th) network, l represents the l-th causal convolutional layer, and the historical traffic sequence is input to obtain the convolutional time characteristics of the base station traffic.

[0083] Next, use RGCN to propagate and aggregate the convolutional temporal features of each base station. Input the temporal features of each similar base station into RGCN in chronological order to obtain the aggregated temporal features of each base station. The RGCN calculation process is similar to the RGCN calculation process in the above-mentioned acquisition of spatial environment features, and will not be repeated here. The aggregated temporal features are then input into the attention mechanism inside the base station, where Transformer is used to capture the temporal features. Optionally, the attention mechanism calculation process in Transformer is as follows:

[0084]

[0085]

[0086] Among them, d k represents the normalization parameter, as well as are all learnable parameters, and h is the h-th head in the multi-head attention mechanism. RGCN As input, that is, the aggregated time characteristics of each base station, the time characteristics of the historical traffic sequence of each base station are finally obtained. t .

[0087] Optionally, obtaining the traffic prediction value of the target base station according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station includes:

[0088] According to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, the traffic prediction value of the target base station is obtained by utilizing the attention fusion mechanism between base stations and a multi-layer perceptron.

[0089] In the embodiment of the present invention, based on the attention fusion part between base stations, the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station are used as input, and the relationship between the base stations is captured through the self-attention mechanism between the base stations, and then the multi-layer perceptron (MLP) is used to output the final prediction value. The self-attention mechanism between base stations can be specifically referred to the above-mentioned Attention function. The calculation process of MLP can be expressed as:

[0090]

[0091] Among them, W1, W2, b1 and b2 represent the learnable parameters of the neural network. The output of MLP is the predicted future base station traffic value. x refers to the output result after splicing the spatial environment characteristics and time characteristics into the self-attention mechanism.

[0092] The flow prediction process in the embodiment of the present invention is as follows: Figure 4 As shown in the figure, the data set is first divided to obtain the historical traffic sequence of the base station, and then the urban knowledge graph is used to model the environmental information around the base station and mine the spatial characteristics of the base station. Then the base station map is extracted from the urban knowledge graph, and the historical traffic sequence is input into the time series modeling part of the system to mine the time characteristics of the base station traffic sequence. The two features are fused through the attention mechanism to obtain the final prediction result.

[0093] Furthermore, the prediction problem of the traffic value of the base station at the next moment can be expressed as follows: Given the historical traffic sequence of the base station (a1, a2, ..., a t-1 ),in The city knowledge graph can be represented as G. The task of base station traffic prediction is to use this known information to predict the traffic value of the base station at the next moment:

[0094]

[0095] The model needs to learn the mapping function F(·). During the model training process, the model's hyperparameters (the dimension of the feature vector, the learning rate, the dimension of the hidden layer of the attention aggregation network, etc.) need to be set. During the network training process, the weights and bias values ​​of each layer of the network can be updated through the Adam method during the back propagation process.

[0096] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the following will compare Figure 5 The specific implementation methods of the present invention are described. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work. In order to make the drawings concise, each figure only schematically shows the parts related to the present invention, and they do not represent the actual structure of the product.

[0097] First, divide the data set. For example, use the Shanghai and Nanjing data sets for testing. Divide the two data sets into training set, validation set, and test set in a ratio of 0.7:0.15:0.15. Based on each training set data and the test set data of the product, K = 12 historical traffic sequences and P = 1 traffic sequence of the prediction time window are obtained by the sliding time window method. That is, the traffic sequence within the prediction time window is predicted by the 12 historical traffic sequences.

[0098] Next, the city knowledge graph is input into the environmental information modeling part to obtain the spatial environmental characteristics of the base station; at the same time, the historical traffic sequence of the base station and the base station map extracted from the city knowledge graph are input into the time series modeling part to obtain the temporal characteristics of the base station traffic sequence. The obtained spatial environmental characteristics of the base station and the temporal characteristics of the traffic sequence are input into the attention fusion mechanism between base stations to obtain the final base station traffic prediction value.

[0099] In the embodiment of the present invention, the relationships between base stations are deeply mined, and four types of relationships between base stations are proposed from the time and space dimensions, namely: distance relationship, similar base station traffic trend relationship, similar base station traffic sequence relationship, and similar POI distribution relationship. These four relationships are added to the city knowledge graph, and Transformer is used to model the time series, which improves the accuracy of base station traffic prediction.

[0100] like Figure 6 As shown, an embodiment of the present invention provides a base station traffic prediction device, including:

[0101] The first acquisition module 601 is used to acquire the spatial environment characteristics of the target base station according to the city knowledge graph, wherein the city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the flow trend relationship between the base stations, the similarity of the flow sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations;

[0102] A second acquisition module 602 is used to acquire the time characteristics of the historical traffic sequence of the target base station according to the city knowledge graph and the historical traffic sequence of the target base station;

[0103] The third acquisition module 603 is used to obtain the traffic prediction value of the target base station according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

[0104] Optionally, the first acquisition module is used to acquire the spatial environment characteristics of the target base station based on a city knowledge graph, a knowledge graph representation learning model, and a multi-relationship graph convolutional neural network.

[0105] Optionally, the first acquisition module includes:

[0106] A first acquisition submodule is used to perform high-order singular value decomposition on fact triples in the city knowledge graph based on the knowledge graph representation learning model to obtain an embedding vector corresponding to the base station in the city knowledge graph, where the base stations in the city knowledge graph include the target base station;

[0107] The second acquisition submodule is used to process the embedded vector based on a multi-relational graph convolutional neural network RGCN to obtain the spatial environment characteristics of the target base station.

[0108] Optionally, the second acquisition module includes:

[0109] A third acquisition submodule is used to acquire at least one similar base station of the target base station according to the city knowledge graph, wherein the distance between the similar base station and the target base station is less than a preset distance, the traffic trends of the similar base station and the target base station correspond to the same type of traffic trends, and the similarity of the traffic sequence between the similar base station and the target base station is greater than a first preset similarity value, and the similarity of the POI distribution between the similar base station and the target base station is greater than a second preset similarity value;

[0110] The fourth acquisition submodule is used to acquire the time characteristics of the historical traffic sequence of the target base station according to the historical traffic sequence of the similar base stations and the historical traffic sequence of the target base station.

[0111] Optionally, the fourth acquisition submodule includes:

[0112] A first acquisition unit is used to perform one-dimensional causal convolution processing on the historical traffic sequences of the target base station and each of the similar base stations to obtain convolution time features of the historical traffic sequences of the target base station and each of the similar base stations;

[0113] A second acquisition unit is used to input the convolution time feature into a multi-relation graph convolutional neural network RGCN to obtain the aggregated time feature of the target base station and each of the similar base stations;

[0114] The third acquisition unit is used to acquire the time characteristics of the historical traffic sequence of the target base station according to the attention mechanism in the Transformer and the aggregated time characteristics.

[0115] Optionally, the third acquisition module is used to obtain the traffic prediction value of the target base station based on the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, using the attention fusion mechanism between base stations and a multi-layer perceptron.

[0116] It should be noted that the device is a device corresponding to the above-mentioned method embodiment. All implementation methods of the above-mentioned method embodiment can be applied to the device embodiment and can achieve the same technical effect, which will not be repeated here.

[0117] like Figure 7 As shown, a base station traffic prediction device according to an embodiment of the present invention includes a processor 710 and a transceiver 720, wherein:

[0118] The processor 710 is used to obtain spatial environmental characteristics of a target base station based on a city knowledge graph, wherein the city knowledge graph represents target relationships between base stations, and the target relationships include distance relationships between base stations, traffic trend relationships between base stations, similarity of traffic sequences between base stations, and similarity of distribution of points of interest (POIs) between base stations; obtain temporal characteristics of a historical traffic sequence of the target base station based on the city knowledge graph and the historical traffic sequence of the target base station; and obtain a traffic prediction value of the target base station based on the spatial environmental characteristics of the target base station and the temporal characteristics of the historical traffic sequence of the target base station.

[0119] It should be noted that the device is a device corresponding to the above-mentioned method embodiment. All implementation methods of the above-mentioned method embodiment can be applied to the device embodiment and can achieve the same technical effect, which will not be repeated here.

[0120] like Figure 8As shown, an embodiment of the present invention also provides a base station traffic prediction device, including: a transceiver 810, a processor 800, a memory 820, and a program or instruction stored on the memory 820 and executable on the processor 800; when the processor 800 executes the program or instruction, the steps of the base station traffic prediction method as described above are implemented.

[0121] The transceiver 810 is used to receive and send data under the control of the processor 800 .

[0122] Among them, Figure 8 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 800 and various circuits of memory represented by memory 820 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 810 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. For different user devices, the user interface 830 may also be an interface capable of externally and internally connecting required devices, and the connected devices include but are not limited to a keypad, a display, a speaker, a microphone, a joystick, and the like.

[0123] The processor 800 is responsible for managing the bus architecture and general processing, and the memory 820 can store data used by the processor 800 when performing operations.

[0124] A readable storage medium according to an embodiment of the present invention stores a program or instruction thereon. When the program or instruction is executed by a processor, the steps in the base station traffic prediction method as described above are implemented and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0125] The processor is the processor in the base station traffic prediction device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0126] It should be further explained that the terminals described in this specification include but are not limited to smart phones, tablet computers, etc., and many of the functional components described are called modules in order to more particularly emphasize the independence of their implementation methods.

[0127] In the embodiment of the present invention, module can be implemented with software so that it can be executed by various types of processors. For example, an executable code module of an identification can include one or more physical or logical blocks of computer instructions, for example, it can be constructed as an object, process or function. Nevertheless, the executable code of the identified module does not need to be physically located together, but can include different instructions stored in different positions, and when these instructions are logically combined together, it constitutes a module and realizes the specified purpose of the module.

[0128] In fact, executable code module can be a single instruction or many instructions, and can even be distributed on a plurality of different code segments, distributed among different programs, and distributed across a plurality of memory devices. Similarly, operating data can be identified in the module, and can be implemented and organized in the data structure of any appropriate type according to any appropriate form. The operating data can be collected as a single data set, or can be distributed in different locations (including on different storage devices), and can only be present on a system or network as an electronic signal at least in part.

[0129] When a module can be implemented by software, considering the level of existing hardware technology, a person skilled in the art can build a corresponding hardware circuit to implement the corresponding function of the module that can be implemented by software without considering the cost. The hardware circuit includes a conventional very large scale integration (VLSI) circuit or gate array and existing semiconductors such as logic chips, transistors, or other discrete components. The module can also be implemented by a programmable hardware device, such as a field programmable gate array, a programmable array logic, a programmable logic device, etc.

[0130] The above exemplary embodiments are described with reference to the accompanying drawings, and many different forms and embodiments are feasible without departing from the spirit and teachings of the present invention. Therefore, the present invention should not be constructed as a limitation of the exemplary embodiments proposed herein. More specifically, these exemplary embodiments are provided so that the present invention will be perfect and complete, and the scope of the present invention will be conveyed to those who are familiar with the technology. In these figures, the component sizes and relative sizes may be exaggerated for clarity. The terms used here are only based on the purpose of describing specific exemplary embodiments and are not intended to be limiting. As used herein, unless the text clearly indicates otherwise, the singular forms "one", "an" and "the" are intended to include these multiple forms. It will be further understood that the terms "including" and / or "comprising" when used in this specification indicate the presence of the features, integers, steps, operations, components and / or components, but do not exclude the presence or increase of one or more other features, integers, steps, operations, components, components and / or their groups. Unless otherwise indicated, when stated, a range of values ​​includes the upper and lower limits of that range and any subranges therebetween.

[0131] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A base station traffic prediction method, characterized in that: include: According to the city knowledge graph, the spatial environment characteristics of the target base station are obtained. The city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the flow trend relationship between the base stations, the similarity of the flow sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations. According to the city knowledge graph and the historical traffic sequence of the target base station, a time feature of the historical traffic sequence of the target base station is obtained; A traffic prediction value of the target base station is obtained according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

2. The method according to claim 1, characterized in that The step of obtaining the spatial environment characteristics of the target base station according to the city knowledge graph includes: The spatial environment characteristics of the target base station are obtained based on the city knowledge graph, the knowledge graph representation learning model and the multi-relationship graph convolutional neural network.

3. The method according to claim 2, characterized in that The obtaining of the spatial environment characteristics of the target base station according to the city knowledge graph, the knowledge graph representation learning network and the multi-relation graph convolutional neural network includes: Based on the knowledge graph representation learning model, high-order singular value decomposition is performed on the fact triples in the city knowledge graph to obtain an embedding vector corresponding to the base station in the city knowledge graph, where the base stations in the city knowledge graph include the target base station; The embedded vector is processed based on a multi-relational graph convolutional neural network RGCN to obtain the spatial environment characteristics of the target base station.

4. The method according to claim 1, characterized in that: The acquiring the time characteristics of the historical traffic sequence of the target base station according to the city knowledge graph and the historical traffic sequence of the target base station includes: According to the city knowledge graph, at least one similar base station of the target base station is obtained, the distance between the similar base station and the target base station is less than a preset distance, the traffic trends of the similar base station and the target base station correspond to the same type of traffic trends, and the similarity of the traffic sequence between the similar base station and the target base station is greater than a first preset similarity value, and the similarity of the POI distribution between the similar base station and the target base station is greater than a second preset similarity value; According to the historical traffic sequence of the similar base stations and the historical traffic sequence of the target base station, a time feature of the historical traffic sequence of the target base station is acquired.

5. The method according to claim 4, characterized in that The acquiring the time characteristics of the historical traffic sequence of the target base station according to the historical traffic sequence of the similar base station and the historical traffic sequence of the target base station includes: Performing one-dimensional causal convolution processing on the historical traffic sequences of the target base station and each of the similar base stations to obtain convolution time features of the historical traffic sequences of the target base station and each of the similar base stations; Inputting the convolution time feature into a multi-relation graph convolutional neural network RGCN to obtain the aggregated time feature of the target base station and each of the similar base stations; According to the attention mechanism in the Transformer and the aggregated time features, the time features of the historical traffic sequence of the target base station are obtained.

6. The method according to claim 1, characterized in that The obtaining, according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, a traffic prediction value of the target base station comprises: According to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station, the traffic prediction value of the target base station is obtained by utilizing the attention fusion mechanism between base stations and a multi-layer perceptron.

7. A base station traffic prediction device, characterized in that: include: The first acquisition module is used to acquire the spatial environment characteristics of the target base station according to the city knowledge graph, wherein the city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the flow trend relationship between the base stations, the similarity of the flow sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations; A second acquisition module is used to acquire the time characteristics of the historical traffic sequence of the target base station according to the city knowledge graph and the historical traffic sequence of the target base station; The third acquisition module is used to obtain the traffic prediction value of the target base station according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

8. A base station traffic prediction device, characterized in that: include: transceivers and processors; The processor is used to obtain the spatial environment characteristics of the target base station according to the city knowledge graph, the city knowledge graph represents the target relationship between the base stations, and the target relationship includes the distance relationship between the base stations, the flow trend relationship between the base stations, the similarity of the flow sequence between the base stations, and the similarity of the distribution of points of interest (POI) between the base stations; According to the city knowledge graph and the historical traffic sequence of the target base station, a time feature of the historical traffic sequence of the target base station is obtained; A traffic prediction value of the target base station is obtained according to the spatial environment characteristics of the target base station and the time characteristics of the historical traffic sequence of the target base station.

9. A base station traffic prediction device, comprising: A transceiver, a processor, a memory, and a program or instruction stored in the memory and executable on the processor; characterized in that when the processor executes the program or instruction, the steps of the base station traffic prediction method as described in any one of claims 1 to 6 are implemented.

10. A readable storage medium having a program or instruction stored thereon, characterized in that: When the program or instruction is executed by the processor, the steps of the base station traffic prediction method as described in any one of claims 1 to 6 are implemented.

Images