A flow card management method and system

Abstract

The present invention discloses a flow card management method and system, which improves the accuracy of flow card feature analysis, thereby improving the accuracy and efficiency of flow card recommendation to users, and further improving the effectiveness of flow card management. Recommending to users the Internet flow demand that fits the user's own needs, that is, improving the accuracy of flow card recommendation, users can choose flow cards with high fit (ranked high in the recommendation list), reducing the waste of flow, saving users' expenses, and reducing the trial and error rate of users in purchasing flow cards, saving users' card selection costs and energy, and improving user experience. At the same time, there is no need to produce redundant flow cards, saving flow card resources, and reducing the number of flow cards, reducing the cost of flow card management for operators, and improving the efficiency and safety of flow card management. In addition, intelligent recommendation of flow cards that meet the needs of users is efficient, low cost, and fast response.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a flow card management method and system. Background Art

[0002] With the popularization of mobile Internet, traffic cards serve as an important carrier of mobile data services. In the era of rapid development of the Internet, the number of Internet users has increased dramatically. Different users have different Internet needs, and use different traffic packages and traffic cards. Operators provide many traffic products, such as traffic cards and traffic packages. Each category will also have different specifications such as data volume, price, and usage time limit. Even with many categories, it is difficult to have a traffic package that is completely suitable for users. For example, users cannot accurately predict how much traffic they will use. Therefore, it is difficult for users to select traffic products that suit them from many traffic packages. At present, users often waste too much traffic on the traffic cards they use, and the traffic packages do not have enough traffic, so they need to pay extra to buy expensive traffic or order additional traffic cards. In summary, the traffic card services currently provided have the following problems: first, there is a waste of traffic resources, second, users spend a lot, and third, traffic card resources are wasted. In addition, the management efficiency of traffic cards directly affects the operating costs and user experience of operators. Traditional traffic card management methods mainly rely on manual monitoring and scheduling, which have problems such as low efficiency, high cost, and slow response. Summary of the invention

[0003] The purpose of the present invention is to provide a flow card management method and system to solve the above problems existing in the prior art.

[0004] An embodiment of the present invention provides a flow card management method, the method comprising:

[0005] Obtaining a flow data set of multiple flow cards to be selected, the flow data set includes multiple usage data, the usage data corresponds to the flow cards to be selected one by one; the usage data includes flow consumption data and user behavior data; the flow consumption data corresponds to the user behavior data; the user behavior data includes an online time series, the online time series includes multiple online time segments; the flow consumption data includes multiple flow consumption amounts; the flow consumption amounts correspond to the online time segments one by one;

[0006] Through the flow card feature extraction model, the flow card feature of each flow card to be selected is obtained based on the flow data set; multiple flow card features are correspondingly obtained for multiple flow cards to be selected;

[0007] Obtain historical traffic usage information of a user traffic card used by the user, the historical traffic usage information including multiple historical time periods and traffic usage corresponding to each historical time period;

[0008] Obtaining user traffic characteristics based on historical traffic usage information;

[0009] Based on the flow characteristics of the user and the characteristics of multiple flow cards, a flow card recommendation list of the user is obtained; the flow card recommendation list includes multiple flow cards to be selected, and the multiple flow cards to be selected are arranged in a recommended order;

[0010] Recommend the traffic card recommendation list to the user;

[0011] Among them, the traffic card feature extraction model includes embedding layer, convolution layer, maximum pooling layer, bidirectional LSTM layer, global maximum pooling layer and fully connected layer; the bidirectional LSTM layer includes forward LSTM and backward LSTM.

[0012] Optionally, the flow card features of each flow card to be selected are obtained based on the flow data set through a flow card feature extraction model, including:

[0013] The flow data set is input into the flow card feature extraction model, and the dimension of the flow data set is adjusted through the embedding layer to obtain the first flow data;

[0014] Performing a convolution operation on the first flow data through a convolution layer to obtain first convolution data;

[0015] Performing a pooling operation on the first convolution data through a maximum pooling layer to obtain maximum pooling data;

[0016] Perform forward prediction on the maximum pooling data through the forward LSTM of the bidirectional LSTM layer to obtain forward prediction data;

[0017] Perform backward prediction on the maximum pooling data through the backward LSTM of the bidirectional LSTM layer to obtain backward prediction data;

[0018] Fuse the forward prediction data with the maximum pooling data to obtain fused forward data;

[0019] Fusing the fused forward data with the backward prediction data to obtain fused prediction data;

[0020] Perform global pooling operation on the fused prediction data through the global maximum pooling layer to obtain global pooling data;

[0021] The global pooled data is input into the fully connected layer, and the fully connected layer outputs the flow card features of each flow card to be selected.

[0022] Optionally, the user's traffic characteristics are obtained based on historical traffic usage information, including:

[0023] Inputting historical traffic usage information into a traffic card feature extraction model to obtain a first feature vector;

[0024] Performing a dimension reduction operation on the flow card features through an embedding layer to obtain a second feature vector; the dimension of the second feature vector is the same as that of the first feature vector;

[0025] Obtain a cross entropy vector between the second eigenvector and the first eigenvector, the cross entropy vector comprising a plurality of cross entropies, each cross entropy being a cross entropy between a first feature and a second feature corresponding to each other, the first feature being an element in the first eigenvector, and the second feature being an element in the second eigenvector;

[0026] Construct a weight adjustment matrix through the cross entropy vector;

[0027] The weight of the flow card feature extraction model is adjusted by using a weight adjustment matrix to obtain an adjusted flow card feature extraction model;

[0028] Input the historical traffic usage information into the adjusted traffic card feature extraction model to obtain the user's traffic characteristics.

[0029] Optionally, the flow card feature extraction model is weighted adjusted by a weight adjustment matrix to obtain an adjusted flow card feature extraction model, including:

[0030] The weights of the embedding layer, convolution layer, maximum pooling layer, global maximum pooling layer and fully connected layer are convolved through the weight adjustment matrix.

[0031] Optionally, a weight adjustment matrix is ​​constructed by using the cross entropy vector, including:

[0032] The elements in the cross entropy vector are used as the diagonal elements of the weight adjustment matrix, and the non-diagonal elements of the weight adjustment matrix are set to 0.

[0033] Optionally, based on the traffic characteristics of the user and the characteristics of multiple traffic cards, obtaining a traffic card recommendation list for the user includes:

[0034] Obtain the relative entropy between each flow card feature and the flow feature, and obtain multiple relative entropies corresponding to multiple flow card features; the relative entropies correspond to the flow cards to be selected one by one;

[0035] Arrange the relative entropy in order from small to large;

[0036] Sort the candidate flow cards corresponding to the relative entropy in the position where the relative entropy is located to obtain a flow card recommendation list.

[0037] Optionally, recommend a traffic card recommendation list to the user, including:

[0038] Prioritize the push of candidate traffic cards with small relative entropy to users.

[0039] Optionally, the method further includes:

[0040] If trigger information is received that the user selects a candidate flow card in the flow card recommendation list, the candidate flow card selected by the user is deleted from the flow card recommendation list; and the flow card recommendation list is updated.

[0041] The embodiment of the present invention further provides a flow card management system, the system comprising:

[0042] The acquisition module is used to obtain the flow data sets of multiple flow cards to be selected, the flow data sets include multiple usage data, the usage data correspond to the flow cards to be selected one by one; the usage data include flow consumption data, user behavior data, and signal strength; the flow consumption data correspond to the user behavior data; the user behavior data includes an online time series, and the online time series includes multiple online time segments; the flow consumption data includes multiple flow consumption amounts; the flow consumption amounts correspond to the online time segments one by one;

[0043] The analysis module is used to obtain the flow card features of each flow card to be selected based on the flow data set through the flow card feature extraction model; obtain multiple flow card features corresponding to multiple flow cards to be selected; obtain historical flow usage information of the user flow card used by the user, the historical flow usage information includes the historical time period and the flow usage amount corresponding to the historical time period; obtain the user's flow features based on the historical flow usage information; obtain the user's flow card recommendation list based on the user's flow features and multiple flow card features; the flow card recommendation list contains multiple flow cards to be selected, and the multiple flow cards to be selected are arranged in the order of recommendation;

[0044] The management module is used to recommend the traffic card recommendation list to the user.

[0045] Among them, the traffic card feature extraction model includes embedding layer, convolution layer, maximum pooling layer, bidirectional LSTM layer, global maximum pooling layer and fully connected layer; the bidirectional LSTM layer includes forward LSTM and backward LSTM.

[0046] Optionally, the management module is further used to: if trigger information is received that the user selects a candidate flow card in the flow card recommendation list, delete the candidate flow card selected by the user from the flow card recommendation list; and update the flow card recommendation list.

[0047] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0048] The embodiment of the present invention provides a flow card management method and system, which improves the accuracy of flow card feature analysis, thereby improving the accuracy and efficiency of recommending flow cards to users, and further improving the effectiveness of flow card management. That is, the flow card recommended to the user through the above scheme fits the user's own Internet traffic demand, that is, the accuracy of flow card recommendation is improved, and the user can choose a flow card with a high degree of fit (ranked high in the recommendation list), which reduces the waste of traffic and saves the user's expenses. At the same time, it reduces the trial and error rate of users in purchasing flow cards, saves the user's card selection cost and energy, and improves the user experience. At the same time, it improves the user's use of flow cards that meet the user's expectations and needs, does not need to produce redundant flow cards, saves flow card resources, and reduces the number of flow cards, reducing the cost, efficiency and security of flow card management by operators. In addition, intelligent recommendation of flow cards that meet the needs of users is efficient, low cost and fast response. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of a flow card management method provided by an embodiment of the present invention.

[0050] Figure 2 It is a flow chart of another flow card management method provided by an embodiment of the present invention.

[0051] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0052] Markings in the figure: 500 - bus; 501 - receiver; 502 - processor; 503 - transmitter; 504 - memory; 505 - bus interface. DETAILED DESCRIPTION

[0053] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0054] Example

[0055] In an embodiment of the present invention, the provided traffic card management method can be used to manage physical traffic cards, can also be used to manage virtual traffic cards, and can also be used to manage traffic packages.

[0056] The embodiment of the present invention provides a flow card management method, such as Figure 1 As shown, the method includes:

[0057] S101: Obtaining traffic data sets of multiple traffic cards to be selected.

[0058] In the embodiment of the present invention, the traffic data set includes multiple usage data, and the usage data corresponds one-to-one to the selected traffic card. The usage data includes traffic consumption data and user behavior data. The usage data may also include signal strength, and the signal strength can be detected by a mobile device (such as a user's mobile phone) and a terminal device (such as a system server). The traffic consumption data corresponds one-to-one to the user behavior data. The user behavior data includes an Internet access time series, and the Internet access time series includes multiple Internet access time segments. The traffic consumption data includes multiple traffic consumption amounts, and the traffic consumption amounts correspond one-to-one to the Internet access time segments. The Internet access time segment is the length of time that the user goes online each time. Optionally, the user's Internet access time is marked in the traffic consumption. The Internet access time includes a start time point and an end time point. The Internet access time segment is equal to the time interval between the end time point and the start time point.

[0059] S102: Obtain the flow card features of each flow card to be selected based on the flow data set through the flow card feature extraction model.

[0060] In an embodiment of the present invention, multiple traffic card features are obtained corresponding to multiple traffic card selections. In an embodiment of the present invention, a traffic card feature extraction model combining a long short-term memory network (Long Short-Term Memory, LSTM) and a convolutional neural network (Convolutional Neural Networks, CNN) is constructed, and the traffic data set is feature extracted by the traffic card feature extraction model. Specifically, the traffic card feature extraction model includes an embedding layer, a convolution layer, a maximum pooling layer, a bidirectional LSTM layer, a global maximum pooling layer and a fully connected layer; the bidirectional LSTM layer includes a forward LSTM and a backward LSTM. Optionally, the specific structure of the traffic card feature extraction model is an embedding layer (Embedding), a convolution layer (Conv1D), a maximum pooling layer (MaxPooling1D), a bidirectional LSTM layer (Bidirectional LSTM), a global maximum pooling layer (GlobalMaxPooling1D) and a bidirectional LSTM-CNN model of a fully connected layer.

[0061] In an embodiment of the present invention, an embedding layer is used to map high-dimensional sparse input data into a low-dimensional dense representation space. The role of the embedding layer is to convert the input data into a dense vector representation, thereby capturing the semantic relationship and similarity between the input data. In summary, in an embodiment of the present invention, the role of the embedding layer is to map high-dimensional sparse input data into a low-dimensional dense continuous representation to extract the semantic information and features of the data, thereby helping the deep learning model to better understand and process the input data. Thereby improving the generalization ability of the model for data of different dimensions, so that the traffic card feature extraction model can process data of different dimensions. For example, the traffic card feature extraction model can process the traffic card features of each selected traffic card based on the traffic data set, and can also obtain the user's traffic features based on the historical traffic usage information through the traffic card feature extraction model. In the specific processing process, for input data of different dimensions, it is necessary to adjust the weights of each neuron in each layer of the model accordingly.

[0062] The one-dimensional convolution layer (Conv1D) used in this embodiment performs convolution operations on one-dimensional data. The main functions of the one-dimensional convolution layer include feature extraction, feature mapping and pattern recognition. Specifically, the one-dimensional convolution layer performs convolution operations on the input sequence by sliding a one-dimensional convolution kernel (filter) to extract local features in the input sequence. In this process, the convolution kernel calculates the weighted sum of each position in the input sequence and generates a new feature sequence. By performing multiple convolution operations on different convolution kernels, the model can learn features of different scales and abstractions, thereby improving the understanding and characterization of the input data. The role of the one-dimensional convolution layer is to extract local features in the input data by performing convolution operations on one-dimensional data, and help the model learn features of different scales and abstractions to improve the understanding and characterization of the input data. It should be noted that the embodiment of the present invention introduces a maximum pooling layer (MaxPooling) and a global pooling layer (GlobalPooling), and the maximum pooling layer (MaxPooling) and the global pooling layer (GlobalPooling) are used to reduce the dimension of the feature map and extract important features. Specifically:

[0063] The max pooling layer slides a fixed-size pooling window in the input (feature map) and selects the maximum value in each window as the output. This method can retain the important features in the input feature map and discard some unimportant information, which plays a role in dimensionality reduction and key feature extraction. This helps the model learn more robust and abstract feature representations, and plays an important role in the accuracy of feature extraction by the model.

[0064] The global pooling layer performs a pooling operation on the input (the entire feature map) to extract the information of the entire feature map into a value or a vector. The global pooling layer averages or maximizes the feature map of each channel to obtain a value representing the entire channel. In other words, through the above operations, the global pooling layer reduces the dimension of the feature map to a fixed size while retaining important feature information. The global pooling layer is often used as the last layer in a convolutional neural network to convert the feature map into a vector of fixed length for easy classification or regression tasks.

[0065] In general, the introduction of the maximum pooling layer (MaxPooling) and the global pooling layer (GlobalPooling) not only extracts the local important features of each candidate flow card in the flow data set, but also extracts the global important features of each candidate flow card in the flow data set, thus improving the accuracy of feature extraction.

[0066] In order to further extract the accuracy of the features of each candidate traffic card in the traffic data set, a bidirectional long short-term memory network (Bidirectional LSTM) is added, that is, a bidirectional LSTM layer (Bidirectional LSTM) is introduced between the maximum pooling layer (MaxPooling1D) and the global maximum pooling layer (GlobalMaxPooling1D). Specifically:

[0067] First, the input sequence data (output of the maximum pooling layer) is passed to the forward LSTM in time steps (step, set step equal to the online time segment). The forward LSTM will gradually process the input data at each time step and update the hidden state and cell state. At each time step, the forward LSTM will consider the current input and the hidden state of the previous time step to capture the timing information in the sequence data.

[0068] At the same time, the input sequence data (output of the maximum pooling layer) is passed to the reverse LSTM in the reverse time step order. The reverse LSTM will gradually process the input data at each time step and update the hidden state and memory unit in the opposite direction. Through back propagation, the reverse LSTM will capture the reverse time sequence information of the input sequence.

[0069] Then, based on the output of the forward LSTM and the output of the reverse LSTM, an overall fusion output is obtained. In this way, the bidirectional LSTM layer effectively utilizes the information in both the forward and reverse directions, helping the model to better understand and process sequence data, taking into account the relationship between the traffic data set including multiple usage data in time and space, and improving the model's modeling ability and prediction accuracy for sequence data.

[0070] The output of the bidirectional LSTM layer is then used as the input of the global maximum pooling layer, and then the input (the entire feature map) is pooled through the global pooling layer. The global pooling layer reduces the dimension of the feature map to a fixed size while retaining important feature information.

[0071] Finally, the high-level features are obtained by combining and mapping the features output by the global maximum pooling layer through the fully connected layer. Specifically, the fully connected layer linearly combines the features extracted by the global maximum pooling layer and performs nonlinear transformation through the activation function to finally obtain the output of the network and output multiple traffic card features.

[0072] Specifically, the flow card features of each selected flow card are obtained based on the flow data set through the flow card feature extraction model, including:

[0073] The traffic data set is input into the traffic card feature extraction model, and the dimension of the traffic data set is adjusted through the embedding layer to obtain the first traffic data. The first traffic data is convolved through the convolution layer to obtain the first convolution data. The first convolution data is pooled through the maximum pooling layer to obtain the maximum pooling data. The forward LSTM of the bidirectional LSTM layer is used to perform forward prediction on the maximum pooling data to obtain the forward prediction data. The backward LSTM of the bidirectional LSTM layer is used to perform backward prediction on the maximum pooling data to obtain the backward prediction data.

[0074] The forward prediction data is fused with the maximum pooling data to obtain the fused forward data; the fused forward data is fused with the backward prediction data to obtain the fused prediction data. By adopting such a scheme, the temporal influence relationship of the traffic data set including multiple usage data is retained, and the spatial influence relationship of the traffic data set including multiple usage data is extracted, thereby improving the prediction accuracy of the model for sequence data.

[0075] Optionally, the forward prediction data is fused with the maximum pooling data to obtain the fused forward data by concatenating the forward prediction data with the maximum pooling data to obtain the fused forward data, that is, the fused forward data = [forward prediction data, maximum pooling data]. The specific method of fusing the fused forward data with the backward prediction data to obtain the fused prediction data may be to fusing and concatenating the fused forward data with the backward prediction data to obtain the fused prediction number, that is, the fused prediction number = [fused forward data, backward prediction data].

[0076] Optionally, the forward prediction data is fused with the maximum pooling data to obtain the fused forward data, which can also be the following method: the forward prediction data is spliced ​​with the maximum pooling data, and then an attention mechanism is used to learn how to assign different weights to each part to obtain the fused forward data. The fused forward data is fused with the backward prediction data to obtain the fused prediction data, which can specifically be: the fused forward data is spliced ​​with the backward prediction data, and then an attention mechanism is used to learn how to assign different weights to each part to obtain the fused prediction data. In this way, the model can pay more attention to the important parts, and the obtained fused prediction data is more focused on representing the important information of each candidate traffic card obtained by the traffic data set.

[0077] Finally, the fusion prediction data is globally pooled through the global maximum pooling layer to obtain global pooling data. The global pooling data is input into the fully connected layer, and the fully connected layer outputs the traffic card features of each candidate traffic card.

[0078] By adopting the above solution, the accuracy of feature extraction is improved, providing an important foundation for the accuracy and effectiveness of subsequent traffic card management.

[0079] S103: Obtain historical traffic usage information of the user traffic card used by the user, wherein the historical traffic usage information includes multiple historical time periods and the traffic usage corresponding to each historical time period.

[0080] S104: Obtaining the user's traffic characteristics based on historical traffic usage information.

[0081] In an embodiment of the present invention, obtaining the user's traffic characteristics based on historical traffic usage information includes:

[0082] The historical traffic usage information is input into the traffic card feature extraction model to obtain the first feature vector.

[0083] The flow card features are subjected to dimensionality reduction operation through the embedding layer to obtain a second eigenvector; the dimension of the second eigenvector is the same as that of the first eigenvector.

[0084] A cross entropy vector of the second eigenvector and the first eigenvector is obtained, the cross entropy vector comprising a plurality of cross entropies, each cross entropy being a cross entropy of a first feature and a second feature corresponding to each other, the first feature being an element in the first eigenvector, and the second feature being an element in the second eigenvector. The positions of the first feature and the second feature correspond to each other, that is, the first feature and the second feature are arranged in the same order in the second eigenvector and the first eigenvector.

[0085] The weight adjustment matrix is ​​constructed by the cross entropy vector. Specifically, the elements in the cross entropy vector are used as the diagonal elements of the weight adjustment matrix, and the values ​​of the non-diagonal elements of the weight adjustment matrix are 0.

[0086] The weight of the traffic card feature extraction model is adjusted through the weight adjustment matrix to obtain the adjusted traffic card feature extraction model. The historical traffic usage information is input into the adjusted traffic card feature extraction model to obtain the user's traffic characteristics.

[0087] Among them, the weight of the traffic card feature extraction model is adjusted through the weight adjustment matrix to obtain the adjusted traffic card feature extraction model, specifically: the weights of the embedding layer, convolution layer, maximum pooling layer, global maximum pooling layer and full connection layer are convolved through the weight adjustment matrix.

[0088] By adopting the above scheme, the transferability of the traffic card feature extraction model is improved, and the accuracy and efficiency of the user's traffic characteristics are also improved.

[0089] S105: Based on the user's traffic characteristics and characteristics of multiple traffic cards, obtain a traffic card recommendation list for the user.

[0090] The traffic card recommendation list includes multiple traffic cards to be selected, and the multiple traffic cards to be selected are arranged in the order of recommendation. Specifically, the relative entropy between each traffic card feature and the traffic feature is obtained, and multiple traffic card features correspond to multiple relative entropies, and the relative entropies correspond to the traffic cards to be selected one by one.

[0091] The relative entropies are arranged in order from small to large, and the candidate flow cards corresponding to the relative entropies are sorted in the position where the relative entropies are located to obtain a flow card recommendation list.

[0092] S106: Recommend the traffic card recommendation list to the user.

[0093] Specifically, the candidate traffic cards with small relative entropy are pushed to the user first.

[0094] By adopting the above scheme, the accuracy of flow card feature analysis is improved, thereby improving the accuracy and efficiency of flow card recommendation to users, and further improving the effectiveness of flow card management. That is, the flow card recommended to users by the above scheme fits the user's own Internet traffic demand, that is, the accuracy of flow card recommendation is improved, and users can choose flow cards with high fit (ranked high in the recommendation list), which reduces the waste of traffic and saves users' expenses. At the same time, it reduces the trial and error rate of users in purchasing flow cards, saves users' card selection costs and energy, and improves user experience. At the same time, it improves the use of flow cards that meet users' expectations and needs, and there is no need to produce excess flow cards, which meets users' needs for flow cards, saves flow card resources, and reduces the number of flow cards, reducing the cost, efficiency and security of flow card management by operators. In addition, intelligent recommendation of flow cards that meet users' needs is efficient, low cost and fast response.

[0095] As an optional implementation scheme, Figure 2 As shown, the method also includes:

[0096] S107: If trigger information is received indicating that the user has selected a candidate flow card in the flow card recommendation list, the candidate flow card selected by the user is deleted from the flow card recommendation list.

[0097] S108: Update the traffic card recommendation list.

[0098] In an embodiment of the present invention, the method for updating the recommended flow card list is as shown in steps S101 to S106 above, that is, the flow card selected by the user is used as the user flow card, that is, the flow card currently being used by the user, and then the flow data sets of multiple flow cards available for selection in the current time period are obtained, and then steps S102 to S106 are executed to obtain a new recommended flow card list.

[0099] In an embodiment of the present invention, the flow card management method can be used to manage flow packages, that is, by replacing the above-mentioned flow card factors with flow packages, the flow packages can be recommended and managed. That is, it can be replaced with a flow package management method, and the flow package management method includes:

[0100] Obtain a traffic data set of multiple traffic packages to be selected, the traffic data set includes multiple usage data, the usage data corresponds to the traffic packages to be selected one by one; the usage data includes traffic consumption data and user behavior data; the traffic consumption data corresponds to the user behavior data; the user behavior data includes an online time series, the online time series includes multiple online time segments; the traffic consumption data includes multiple traffic consumption amounts; the traffic consumption amounts correspond to the online time segments one by one;

[0101] Using the traffic package feature extraction model, the traffic package feature of each traffic package to be selected is obtained based on the traffic data set; multiple traffic package features are obtained corresponding to multiple traffic packages to be selected;

[0102] Obtain historical traffic usage information of a user traffic package used by the user, the historical traffic usage information including multiple historical time periods and traffic usage corresponding to each historical time period;

[0103] Obtaining user traffic characteristics based on historical traffic usage information;

[0104] Based on the traffic characteristics of the user and the characteristics of multiple traffic packages, a traffic package recommendation list for the user is obtained; the traffic package recommendation list includes multiple traffic packages to be selected, and the multiple traffic packages to be selected are arranged in a recommended order;

[0105] Recommend the traffic package recommendation list to the user.

[0106] By adopting the above scheme, the accuracy and reliability of the analysis of users' demand for traffic cards (traffic packages) can be improved. The traffic cards (traffic packages) recommended to users by the above scheme are in line with the users' own Internet traffic needs, that is, the accuracy of traffic card recommendations is improved, and the traffic cards (traffic packages) that best meet the users' needs are intelligently prioritized, with high accuracy, reliability, high efficiency, low cost and fast response.

[0107] Based on the above-mentioned flow card management method, an embodiment of the present invention provides a flow card management system for executing the above-mentioned flow card management method. The flow card management system comprises:

[0108] The acquisition module is used to obtain the flow data sets of multiple flow cards to be selected, the flow data sets include multiple usage data, the usage data correspond to the flow cards to be selected one by one; the usage data include flow consumption data, user behavior data, and signal strength; the flow consumption data correspond to the user behavior data; the user behavior data includes an online time series, and the online time series includes multiple online time segments; the flow consumption data includes multiple flow consumption amounts; the flow consumption amounts correspond to the online time segments one by one;

[0109] The analysis module is used to obtain the flow card features of each flow card to be selected based on the flow data set through the flow card feature extraction model; obtain multiple flow card features corresponding to multiple flow cards to be selected; obtain historical flow usage information of the user flow card used by the user, the historical flow usage information includes the historical time period and the flow usage amount corresponding to the historical time period; obtain the user's flow features based on the historical flow usage information; obtain the user's flow card recommendation list based on the user's flow features and multiple flow card features; the flow card recommendation list contains multiple flow cards to be selected, and the multiple flow cards to be selected are arranged in the order of recommendation;

[0110] The management module is used to recommend the traffic card recommendation list to the user.

[0111] Among them, the traffic card feature extraction model includes embedding layer, convolution layer, maximum pooling layer, bidirectional LSTM layer, global maximum pooling layer and fully connected layer; the bidirectional LSTM layer includes forward LSTM and backward LSTM.

[0112] Optionally, the management module is further used to: if trigger information is received that the user selects a candidate flow card in the flow card recommendation list, delete the candidate flow card selected by the user from the flow card recommendation list; and update the flow card recommendation list.

[0113] The specific implementation methods of the functions of each module of the above system are as described in the above traffic card management method and will not be repeated here.

[0114] The embodiment of the present invention also provides an electronic device for integrating the above-mentioned flow card management system, such as Figure 3 As shown, the electronic device includes a memory 504, a processor 502, and a computer program stored in the memory 504 and executable on the processor 502. When the processor 502 executes the program, the steps of any of the traffic card management methods described above are implemented.

[0115] Among them, Figure 3 In the embodiment of the present invention, a bus architecture (represented by bus 500) is shown, which may include any number of interconnected buses and bridges, and bus 500 links various circuits including one or more processors represented by processor 502 and memory represented by memory 504. Bus 500 may also link 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. Bus interface 505 provides an interface between bus 500 and receiver 501 and transmitter 503. Receiver 501 and transmitter 503 may be the same element, namely a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 502 is responsible for managing bus 500 and general processing, while memory 504 may be used to store data used by processor 502 when performing operations.

[0116] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of any method of the traffic card management method described above are implemented.

[0117] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0118] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0119] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0120] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0121] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0122] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components in the apparatus according to an embodiment of the present invention. The present invention may also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium, or may have the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

Claims

1. A flow card management method, characterized in that: The method comprises: Obtaining a flow data set of multiple flow cards to be selected, the flow data set includes multiple usage data, the usage data corresponds to the flow cards to be selected one by one; the usage data includes flow consumption data and user behavior data; the flow consumption data corresponds to the user behavior data; the user behavior data includes an online time series, the online time series includes multiple online time segments; the flow consumption data includes multiple flow consumption amounts; the flow consumption amounts correspond to the online time segments one by one; Through the flow card feature extraction model, the flow card feature of each flow card to be selected is obtained based on the flow data set; multiple flow card features are correspondingly obtained for multiple flow cards to be selected; Obtain historical traffic usage information of a user traffic card used by the user, the historical traffic usage information including multiple historical time periods and traffic usage corresponding to each historical time period; Obtaining user traffic characteristics based on historical traffic usage information; Based on the flow characteristics of the user and the characteristics of multiple flow cards, a flow card recommendation list of the user is obtained; the flow card recommendation list includes multiple flow cards to be selected, and the multiple flow cards to be selected are arranged in a recommended order; Recommend the traffic card recommendation list to the user; The traffic card feature extraction model includes an embedding layer, a convolution layer, a maximum pooling layer, a bidirectional LSTM layer, a global maximum pooling layer, and a fully connected layer; the bidirectional LSTM layer includes a forward LSTM and a backward LSTM; based on the traffic data set, the traffic card features of each selected traffic card are obtained, including: The flow data set is input into the flow card feature extraction model, and the dimension of the flow data set is adjusted through the embedding layer to obtain the first flow data; Performing a convolution operation on the first flow data through a convolution layer to obtain first convolution data; Performing a pooling operation on the first convolution data through a maximum pooling layer to obtain maximum pooling data; Perform forward prediction on the maximum pooling data through the forward LSTM of the bidirectional LSTM layer to obtain forward prediction data; The maximum pooling data is backward predicted through the backward LSTM of the bidirectional LSTM layer to obtain the backward prediction data; the forward prediction data is fused with the maximum pooling data to obtain the fused forward data; Fusing the fused forward data with the backward prediction data to obtain fused prediction data; Perform global pooling operation on the fused prediction data through the global maximum pooling layer to obtain global pooling data; The global pooled data is input into the fully connected layer, and the fully connected layer outputs the flow card features of each flow card to be selected.

2. The flow card management method according to claim 1, characterized in that: Obtain the user's traffic characteristics based on historical traffic usage information, including: Inputting historical traffic usage information into a traffic card feature extraction model to obtain a first feature vector; Performing a dimension reduction operation on the flow card features through an embedding layer to obtain a second feature vector; the dimension of the second feature vector is the same as that of the first feature vector; Obtain a cross entropy vector between the second eigenvector and the first eigenvector, the cross entropy vector comprising a plurality of cross entropies, each cross entropy being a cross entropy between a first feature and a second feature corresponding to each other, the first feature being an element in the first eigenvector, and the second feature being an element in the second eigenvector; Construct a weight adjustment matrix through the cross entropy vector; The weight of the flow card feature extraction model is adjusted by using a weight adjustment matrix to obtain an adjusted flow card feature extraction model; Input the historical traffic usage information into the adjusted traffic card feature extraction model to obtain the user's traffic characteristics.

3. The flow card management method according to claim 2, characterized in that: The weight of the flow card feature extraction model is adjusted through the weight adjustment matrix to obtain the adjusted flow card feature extraction model, including: The weights of the embedding layer, convolution layer, maximum pooling layer, global maximum pooling layer and fully connected layer are convolved through the weight adjustment matrix.

4. The flow card management method according to claim 3, characterized in that: The weight adjustment matrix is ​​constructed by the cross entropy vector, including: The elements in the cross entropy vector are used as the diagonal elements of the weight adjustment matrix, and the non-diagonal elements of the weight adjustment matrix are set to 0.

5. The flow card management method according to claim 1, characterized in that: Based on the traffic characteristics of the user and the characteristics of multiple traffic cards, obtaining a traffic card recommendation list for the user includes: Obtain the relative entropy between each flow card feature and the flow feature, and obtain multiple relative entropies corresponding to multiple flow card features; the relative entropies correspond to the flow cards to be selected one by one; Arrange the relative entropy in order from small to large; Sort the candidate flow cards corresponding to the relative entropy in the position where the relative entropy is located to obtain a flow card recommendation list.

6. The flow card management method according to claim 5, characterized in that: Recommend the traffic card recommendation list to the user, including: Prioritize the push of candidate traffic cards with small relative entropy to users.

7. The flow card management method according to claim 1, characterized in that: The method further comprises: If trigger information is received that the user selects a candidate flow card in the flow card recommendation list, the candidate flow card selected by the user is deleted from the flow card recommendation list; and the flow card recommendation list is updated.

8. A flow card management system, characterized in that: The system comprises: An acquisition module is used to obtain a traffic data set of multiple traffic cards to be selected, the traffic data set includes multiple usage data, and the usage data corresponds to the traffic cards to be selected one by one; the usage data includes traffic consumption data, user behavior data, and signal strength; the traffic consumption data corresponds to the user behavior data; the user behavior data includes an Internet access time series, and the Internet access time series includes multiple Internet access time segments; the traffic consumption data includes multiple traffic consumption amounts; the traffic consumption amounts correspond to the Internet access time segments one by one; an analysis module is used to obtain the traffic card features of each traffic card to be selected based on the traffic data set through a traffic card feature extraction model; multiple traffic card features are obtained corresponding to multiple traffic cards to be selected; historical traffic usage information of a user traffic card used by a user is obtained, and the historical traffic usage information includes a historical time period and the traffic usage amount corresponding to the historical time period; the traffic characteristics of the user are obtained based on the historical traffic usage information; based on the traffic characteristics of the user and the characteristics of multiple traffic cards, a traffic card recommendation list of the user is obtained; the traffic card recommendation list contains multiple traffic cards to be selected, and the multiple traffic cards to be selected are arranged in the order of recommendation; A management module is used to recommend a traffic card recommendation list to the user; Among them, the traffic card feature extraction model includes an embedding layer, a convolution layer, a maximum pooling layer, a bidirectional LSTM layer, a global maximum pooling layer, and a fully connected layer; the bidirectional LSTM layer includes a forward LSTM and a backward LSTM; Based on the traffic data set, the traffic card features of each candidate traffic card are obtained, including: The flow data set is input into the flow card feature extraction model, and the dimension of the flow data set is adjusted through the embedding layer to obtain the first flow data; Performing a convolution operation on the first flow data through a convolution layer to obtain first convolution data; Performing a pooling operation on the first convolution data through a maximum pooling layer to obtain maximum pooling data; Perform forward prediction on the maximum pooling data through the forward LSTM of the bidirectional LSTM layer to obtain forward prediction data; Perform backward prediction on the maximum pooling data through the backward LSTM of the bidirectional LSTM layer to obtain backward prediction data; Fuse the forward prediction data with the maximum pooling data to obtain fused forward data; Fusing the fused forward data with the backward prediction data to obtain fused prediction data; Perform global pooling operation on the fused prediction data through the global maximum pooling layer to obtain global pooling data; The global pooled data is input into the fully connected layer, and the fully connected layer outputs the flow card features of each flow card to be selected.

9. The flow card management system according to claim 8, characterized in that: The management module is also used for: if trigger information is received that the user selects a candidate flow card in the flow card recommendation list, deleting the candidate flow card selected by the user from the flow card recommendation list; and updating the flow card recommendation list.

Images