Data processing method and device, equipment, medium and product

Abstract

The invention discloses a data processing method and device, equipment, a medium and a product. The method comprises the following steps: acquiring a user data set and a network data set; inputting the user data set and the network data set into a multi-modal deep learning model based on a restricted Boltzmann machine (RBM), and clustering feature vectors output by a hidden layer of the multi-modal deep learning model based on the RBM to obtain a clustering portrait of the user and the network; determining a first network resident factor of the user according to the clustering portrait of the user and the network; a processing strategy is determined according to the factor that the user resides in the first network, and the processing strategy comprises the steps of expanding the coverage range of the second network and generating reminding information. Through the technical scheme of the invention, it can be ensured that the user resides in the second network as much as possible, the wireless resource utilization rate of the area to be subjected to frequency reduction network quit is improved, and meanwhile, the user experience is improved. And the user experience in the region is ensured.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of communication, and in particular to a data processing method, device, equipment, medium and product. BACKGROUND

[0002] With the rapid development of communication technology, the popularity of 5G network makes 2G network gradually face the realistic demand of network withdrawal. At present, the mobile communication network contains 2 / 4 / 5G multi-layer network, different users may stay in different types of network, and the operator needs to maintain multiple layers of network, thereby there are problems of high network maintenance cost, low resource utilization, and large network security hidden danger. SUMMARY

[0003] Embodiments of the present application provide a data processing method, device, equipment, medium and product to solve at least one of the above problems.

[0004] According to an aspect of the present application, a data processing method is provided, comprising:

[0005] obtaining a user data set and a network data set;

[0006] inputting the user data set and the network data set into a multi-modal deep learning model based on a restricted Boltzmann machine (RBM), clustering feature vectors output by a hidden layer of the multi-modal deep learning model based on RBM to obtain a clustering portrait of users and networks;

[0007] determining a user staying in a first network factor according to the clustering portrait of users and networks, wherein the user staying in the first network factor includes at least one of a second network coverage factor, a terminal adaptation factor, a user intention factor and a value-added service factor;

[0008] determining a processing strategy according to the user staying in the first network factor, wherein the processing strategy includes expanding the second network coverage range and generating a reminder information.

[0009] According to another aspect of the present application, a data processing device is provided, comprising:

[0010] an obtaining module configured to obtain a user data set and a network data set;

[0011] a clustering portrait of users and networks determination module configured to input the user data set and the network data set into a multi-modal deep learning model based on a restricted Boltzmann machine (RBM), and cluster feature vectors output by a hidden layer of the multi-modal deep learning model based on RBM to obtain a clustering portrait of users and networks;

[0012] a user stays in a first network factor determination module configured to determine a user stays in a first network factor according to the cluster portrait of the user and the network, wherein the user stays in the first network factor comprises at least one of a second network coverage factor, a terminal adaptation factor, a user intention factor, and a value-added service factor;

[0013] a processing strategy determination module configured to determine a processing strategy according to the user stays in the first network factor, wherein the processing strategy comprises expanding the second network coverage range and generating reminder information.

[0014] According to another aspect of the present application, an electronic device is provided, the electronic device comprising:

[0015] at least one processor; and

[0016] a memory connected to the at least one processor in communication; wherein

[0017] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the data processing method according to any one of the embodiments of the present application.

[0018] According to another aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium stores computer instructions for enabling a processor to implement the data processing method according to any one of the embodiments of the present application when executed by the processor.

[0019] According to another aspect of the present application, a computer program product is provided, the computer program is executed by a processor to implement the data processing method according to any one of the embodiments of the present application.

[0020] The embodiments of the present application obtain a user data set and a network data set; input the user data set and the network data set into a multi-modal deep learning model based on a restricted Boltzmann machine (RBM), cluster feature vectors output by a hidden layer of the multi-modal deep learning model based on the RBM to obtain a cluster portrait of the user and the network; determine a user stays in a first network factor according to the cluster portrait of the user and the network, wherein the user stays in the first network factor comprises at least one of a second network coverage factor, a terminal adaptation factor, a user intention factor, and a value-added service factor; determine a processing strategy according to the user stays in the first network factor, wherein the processing strategy comprises expanding the second network coverage range and generating reminder information, determine the processing strategy according to the user data set and the network data set, and try to ensure that the user stays in the second network, that is, an efficient network, which can improve the utilization rate of wireless resources in the area to be reduced in frequency and off-network, while ensuring the user experience in the area, so that the work of reducing frequency and off-network is more orderly and reasonable.

[0021] It should be understood that the matters described in this section are not intended to identify key or critical elements of the embodiments of the present application or to delineate the scope of the present application. Other features of the present application will be apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as limiting the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0023] Figure 1 is a flow chart of a data processing method in the embodiments of the present application;

[0024] Figure 2 is a structural schematic diagram of a neural network architecture in the embodiments of the present application;

[0025] Figure 3 is a topological structure schematic diagram of a BP neural network in the embodiments of the present application;

[0026] Figure 4 is a structural schematic diagram of a multi-modal deep learning model based on a restricted Boltzmann machine in the embodiments of the present application;

[0027] Figure 5 is a structural schematic diagram of a data processing device in the embodiments of the present application;

[0028] Figure 6 is a structural schematic diagram of an electronic device in the embodiments of the present application. DETAILED DESCRIPTION

[0029] In order to make the person skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0031] It can be understood that, before using the technical solutions disclosed in the embodiments of the present disclosure, the type, use range, use scenario, etc. of the personal information involved in the present disclosure should be informed to the user and the authorization of the user should be obtained in accordance with relevant laws and regulations.

[0032] Embodiment one

[0033] Figure 1 A flowchart of a data processing method provided by the embodiment of the present application, the embodiment can be applicable to the case of data processing, the method can be executed by the data processing device in the embodiment of the present application, and the device can be realized in the form of software and / or hardware, as shown in the figure, the method specifically includes the following steps: Figure 1

[0034] S110, obtaining a user data set and a network data set.

[0035] In the embodiment, the user data set and the network data set are collected and prepared, the user data collection includes but is not limited to user traffic type, traffic volume, use frequency, terminal type, average revenue per user (ARPU) value, service detail data, complaint data, etc.; the network data set includes but is not limited to base station 4 / 5G engineering parameters, resident user model accessing the base station, user number, measurement report (MR) reported by the user, network coverage rate and quality of the cell, MDT network coverage map, etc. The data preparation stage includes data acquisition, data preprocessing, data standardization and normalization steps. The data acquisition stage uses network management system, MDT workbench and other systems to collect the required data. In the preprocessing stage, the collected text data is discarded, the non-numeric (Not a Number, NaN) value is discarded, the special character is deleted, the format is unified, the image data is unified in format and resolution. In the standardization and normalization stage, the text data is vectorized and the image pixel vector is normalized by using the maximum and minimum values. A complete data set that can be used for further analysis and training is obtained after processing. ​

[0036] Optionally, the user data set comprises at least one of user traffic type, traffic volume, network usage frequency, terminal type, user average revenue, service detail data and complaint data, and the network data set comprises at least one of base station operating parameter, resident user type accessing the base station, user number, measurement report, network coverage rate of a cell, network coverage quality of the cell and minimization of drive tests (MDT) network coverage map.

[0037] In the embodiment, the user average revenue (ARPU) = total revenue in a specific time period ÷ average user number in the time period. The measurement report is a measurement report (MR) reported by a user. The minimization of drive tests (MDT) network coverage map is a map for showing wireless network coverage conditions, which is drawn by means of data collected by MDT technology. The service detail data is structured data for recording detailed interaction behavior, consumption record or operation track of a user in a specific service scenario, and usually contains specific information such as time, behavior type, amount, object, etc. The user number is the resident user number accessing the base station. The base station operating parameter can be a base station 4 / 5G operating parameter. The terminal type can be a mobile terminal type used by a user, and the terminal type includes 5G supported mobile phone and 5G unsupported mobile phone. The complaint data can be complaint data of a user for a 5G network. The network usage frequency is the frequency of using each network by a user.

[0038] In S120, the user data set and the network data set are input into a multi-modal deep learning model based on a restricted Boltzmann machine (RBM), and a feature vector output by a hidden layer of the multi-modal deep learning model based on the RBM is clustered to obtain a clustered portrait of a user and a network.

[0039] Optionally, the clustering of the feature vector output by the hidden layer of the multi-modal deep learning model based on the RBM to obtain the clustered portrait of the user and the network comprises:

[0040] The feature vector output by the hidden layer of the multi-modal deep learning model based on the RBM is subjected to dimension compression to obtain a dimension-compressed feature vector.

[0041] In the embodiment, the manner of dimension-compressing the feature vector output by the hidden layer of the multi-modal deep learning model based on the RBM to obtain the dimension-compressed feature vector can be that the feature vector output by the hidden layer of the multi-modal deep learning model based on the RBM is subjected to dimension compression based on a principal component analysis algorithm to obtain the dimension-compressed feature vector.

[0042] Determine the target clustering algorithm according to the distribution characteristics of the dimensionally compressed feature vector.

[0043] In this embodiment, the manner of determining the target clustering algorithm according to the distribution characteristics of the dimensionally compressed feature vector can be: if the distribution characteristics of the dimensionally compressed feature vector are spherical or uniform distribution, determining the target clustering algorithm as K-means algorithm or K-medoids algorithm; if the distribution characteristics of the dimensionally compressed feature vector are non-spherical or complex distribution, determining the target clustering algorithm as DBSCAN algorithm or spectral clustering algorithm.

[0044] In this embodiment, the target clustering algorithm includes K-Means algorithm, DBSCAN algorithm and hierarchical clustering algorithm. The K-Means algorithm is suitable for scenarios where the feature distribution is relatively uniform and the number of clusters can be preset (such as clusters known to be divided into "highly active users-strong coverage network" and "low active users-weak coverage network"), has high computational efficiency, and is suitable for large-scale feature vectors. The DBSCAN algorithm does not need to preset the number of clusters, can automatically identify noise points (such as abnormal user behavior characteristics), and is suitable for scenarios where there are outliers (such as the association characteristics of burst high traffic users and base stations). The hierarchical clustering algorithm displays the hierarchical relationship between clusters by constructing a tree structure (such as "4G resident users" can be divided into "VoLTE dependent type" and "traffic intensive type" sub-clusters), and is suitable for scenarios that need to mine multi-modal feature hierarchical association.

[0045] Clustering the dimensionally compressed feature vector based on the target clustering algorithm to obtain a clustering portrait of users and networks.

[0046] Optionally, the multi-modal deep learning model based on RBM includes a visible layer and a hidden layer, and each of the visible layer and the hidden layer includes a plurality of RBM sub-networks, each RBM sub-network of the visible layer is connected to all RBM sub-networks of the hidden layer, and each RBM sub-network of the hidden layer is connected to all RBM sub-networks of the visible layer.

[0047] Optionally, the training process of the multi-modal deep learning model based on RBM includes:

[0048] Determine the activation probability of each RBM sub-network based on the first formula, and the first formula is:

[0049]

[0050] wherein, is the activation probability of the i-th RBM sub-network of the L+1-th layer, is the connection weight between the j-th RBM sub-network of the L-th layer and the i-th RBM sub-network of the L+1-th layer, a state of the jth RBM subnetwork of the Lth layer;

[0051] extracting each single-modal feature according to the activation probability of each RBM subnetwork;

[0052] determining a joint probability according to each single-modal feature;

[0053] unsupervised training the to-be-trained model based on the joint probability, to obtain an RBM-based multi-modal deep learning model.

[0054] At present, deep learning models have been successfully applied to more and more single-modal detection. Deep learning models have the ability to learn feature representations directly from raw data (text, speech, images) rather than relying on domain-specific features and their hierarchical structures. Neural network architectures such as Figure 2 are shown in the drawings. This architecture is composed of interconnected units, which are grouped together by layers. Data information propagates through the layers, and each layer transforms its input according to the internal feature representation, thereby making a global contribution to the final classification result. The first layer (input layer) accepts raw or lightly processed format data, while the last layer (output layer) provides classes (clustering results of users and networks) according to the values of the relevant units. The layers between the input layer and the output layer are called hidden layers, and the unit parameters are automatically determined during the training process, depending on the type of layer and the training method, pre-training and fine-tuning.

[0055] To adjust the network parameters to match the expected values observed from the training set, the training set is used to train the network and the error correction algorithm is used to correct the error. A multi-layer feedforward neural network using error correction is called a BP neural network. During the training process, a gradient search algorithm is used to minimize the mean square error between the network output and the expected value. The topology of the BP neural network is shown in Figure 3 .

[0056] Suppose Figure 3 the number of input nodes of the network is n, the total number of hidden nodes is δ, the total number of output nodes is m, the weight between the input layer and the hidden layer is C, the weight between the hidden layer and the output layer is W, and the activation functions of the hidden layer and the output layer are f1(·) and f2(·) respectively. The output of the hidden node and the output of the output node are represented as:

[0057]

[0058] where z k is the output of the kth hidden node, c kσ is the weight parameter (connecting the input x σ and the kth hidden node), x σ is the σth input feature, y ψ is the output of the ψth output node, and wψk is the connection weight between the th output node and the (k-1)th hidden node in the upper layer.

[0059] For the st sample, the output of the network is The desired output is The mean square error is:

[0060]

[0061] For all S training samples, the global mean square error is:

[0062]

[0063] Set the learning rate η, and use the back propagation of cumulative error to adjust the weights between the output layer and the hidden layer, so as to minimize the global mean square error:

[0064]

[0065] Due to the feature representation learning of the deep learning architecture, it is not necessary to extract specific field features from the input data in the preprocessing stage, which has two important influences on deep learning data. On the one hand, it can train custom layer representations to combine information from different single modalities and cross modalities. On the other hand, the hierarchical structure in the learning representation means that cross-modal relationships can be learned at low and high abstraction levels. The multi-layer feature representation of the deep learning architecture allows them to extract more complex information than commonly used shallow methods.

[0066] Each neural network (i.e. BP network) in the multi-modal deep learning model has a separate architecture branch, and there is no internal branch connection between the layers until the last multi-modal architecture unifies the information from different neural networks. Each independent BP neural network learns a representation related to a single modality, while multi-modal deep learning has powerful feature extraction and fusion capabilities, breaking the limitations of single modality information for user and network analysis. In order to take advantage of the multi-modal deep learning model in feature fusion, the present application proposes a multi-modal deep learning model based on restricted Boltzmann machine (RBM) (as shown in Figure 4 for user and base station analysis clustering in a system with multiple data.

[0067] As an unsupervised probabilistic graphical model, RBM can learn cross-modal joint representations without labeled data, model the joint probability distribution between multi-modal data, and realize cross-modal information fusion through shared hidden layer features, and is suitable for processing large-scale unlabeled multi-modal data. RBM is composed of two layers of visible layer and hidden layer, with no connection within the layer and full connection between the layers. The multi-modal deep learning model based on RBM:

[0068] Each individual RBM unit within layer (L+1) of state (Where w is the unit weight connecting the jth node in layer L to the ith node in layer L+1. In this fully connected structure, each node has a connection to all nodes on layer L with weights

[0069] As shown in Figure 4 , there are separate architecture branches for each data modality, with no connections between these initial layers until the independent branch data is unified in the final output layer of the architecture. Thus, all layers contribute to learning a joint representation of all data. This is represented specifically as:

[0070]

[0071] where v represents the input modality data, containing user and base station dataset information, h represents the hidden units of the network hidden layers, Z(Θ) is a normalization function, ε is the cumulative state of the final output layer, and Θ = {a, W} represents the parameter set of the RBM (a is the bias of the hidden layer). Essentially, the feature learning is performed at the network parameter level, represented by the weights between nodes and the depth of the network. Training of the multi-modal network is performed through backpropagation, running several times over the training set and gradually adjusting the network parameters to match the expectations observed from the training set. The present application uses the user and network dataset to train the RBM multi-modal deep learning model, realizing the clustering portrait of the user and the network.

[0072] Optionally, according to the respective single-modal features, a joint probability is determined, including:

[0073] Feature aggregation is performed on the respective single-modal features to obtain a joint abstract feature;

[0074] Based on a second formula, a joint probability is determined, the second formula being:

[0075]

[0076] where v includes the user dataset and the network dataset, h is the joint abstract feature, Θ is the RBM parameter, P(v, h; Θ) is the joint probability of v and h, Z(Θ) is the partition function, and ε(v, h; Θ) is the energy function.

[0077] In this embodiment, the partition function makes the probability sum of all possible (v, h) combinations equal to 1.

[0078] In this embodiment, ε(v, h; Θ) = -a T v-b T v-v TWh, a, b are visible layer, hidden layer bias, W is inter-layer weight, the smaller ε(v, h; Θ) is, the more the (v, h) combination conforms to the real distribution of data.

[0079] In the embodiment, the unsupervised training is used to perform unsupervised training on the to-be-trained model, and a multi-modal deep learning model based on RBM is obtained. The core of the unsupervised training is to optimize the joint probability fitting to the real distribution. The specific process is as follows:

[0080] Positive phase (fitting real data): input the training data into the to-be-trained model, calculate the activation probability of the hidden layer, sample the hidden layer state based on the activation probability, and record the joint probability of "real data to hidden layer". Negative phase (fitting model generated data): based on the data distribution simulated by the model generated based on the training data, the training data is sampled based on the activation probability of the hidden layer, the sampling data is obtained, and the visible layer is reconstructed from the sampling data to obtain the reconstructed data, and then the joint probability of "model generated data to hidden layer" is calculated. Parameter update: calculate the gradient difference of the positive phase and the negative phase, and update the RBM parameter Θ={W, a, b}.

[0081] In the embodiment, after the training of a single RBM is completed, the hidden layer thereof can be used as a new visible layer to stack the next layer of RBM. The joint probability of each layer of RBM provides more abstract multi-modal features for the upper layer.

[0082] S130, determining that the user resides in a first network factor according to the clustering portrait of the user and the network.

[0083] The user resides in the first network factor includes at least one of a second network coverage factor, a terminal adaptation factor, a user intention factor, and a value-added service factor.

[0084] In the embodiment, the first network can be an inefficient network, for example, the first network can include a 2G and / or 4G network. The second network can be an efficient network, for example, the second network can include a 5G network.

[0085] In the embodiment, the second network coverage factor can be the second network coverage degree (the second network coverage factor includes: no coverage, weak coverage, and poor quality), the user intention factor includes the intention of whether the user wants to use the second network. The value-added service can include the package fee selected by the user. The terminal adaptation factor can include: machine network mismatch, 5G switch off, and low 5G terminal popularization rate.

[0086] In the embodiment, based on the classification portrait of the user and the network, the main reasons for the terminal user to reside in the 2G and 4G network are analyzed and divided into four categories, one is network coverage, two is terminal mismatch, three is user subjective behavior influence, and four is the influence of user selected package fee, as shown in Table 1:

[0087] Table 1

[0088]

[0089]

[0090] S140, determining a processing strategy according to the user residing in the first network factor.

[0091] The processing strategy comprises: expanding the second network coverage range and generating reminding information.

[0092] In the embodiment, the reminding information comprises at least one of the following: reminding to replace the mobile phone, providing to turn on the 5G switch, and reminding to replace the package fee.

[0093] In the embodiment, targeted solutions are given for different user and network scenarios, as shown in Table 2:

[0094] Table 2

[0095]

[0096]

[0097] According to the results of user clustering, targeted guidance measures and network strategies are proposed for different user groups, as shown in Table 3.

[0098] Table 3

[0099]

[0100]

[0101] The embodiment of the present application relies on the BP neural network to construct a multi-modal deep learning model to perform clustering analysis on user data and network data, and formulates targeted user guidance and network clean-up and optimization upgrade strategies for different user clusters and network scenarios. The clean-up of inefficient networks (such as 2G networks) helps to reduce network operating costs and improve profitability. The redistribution of released 900M and other frequency bands to supplement 4 / 5G frequency bands can expand transmission bandwidth and improve transmission rate and network capacity. Guiding users and optimizing and upgrading networks can help improve user network experience and enterprise brand competitiveness. In poor quality and weak coverage network areas, users may fall back to 2G networks due to poor signal, affecting call quality. By supplementing network sites to improve 4 / 5G network coverage quality, users can use VoLTE / VoNR high-definition voice services. By cleaning up inefficient networks such as 2G networks, network green and sustainable development and recycling of infrastructure resources such as spectrum and site can be promoted. With more than 95% of 4G / 5G users, the economic benefits of retaining old networks such as 2G are significantly lower than the maintenance costs, and network withdrawal is the result of market selection. In addition, it is imperative to optimize and upgrade 4 / 5G networks.

[0102] The embodiment of the present application mainly faces complex and diverse users and network situations, collects and processes relevant information of users and base station networks, constructs a multi-modal deep learning model based on BP network to cluster and profile users and network situations, and formulates corresponding user guidance measures and network strategies for different user clusters and network classifications. Ultimately, while ensuring the improvement of user network experience, the network is cleaned up and upgraded, and the network benefit is improved.

[0103] The embodiment of the present application improves the utilization rate of wireless resources in the area to be reduced in frequency and retires the network, while ensuring the user experience in the area, making the frequency reduction and network retirement work more orderly and reasonable.

[0104] The present application relies on the BP neural network to construct a multi-modal deep learning model to comprehensively analyze user data and network data. First, the data acquisition module acquires user data and network related data, and makes a data set according to the collected data, then transmits the data set to the model for feature representation, finally inputs the feature data of multiple modalities into the multi-modal deep learning model based on the restricted Boltzmann machine for feature extraction and fusion, and outputs the clustered user and network data. Relying on the clustering information to analyze the reasons for the user staying in the inefficient network, and formulating targeted user guidance and network clean-up and optimization upgrade strategies for different user clusters and network scenarios. The method of the present application can significantly improve the user and network checking efficiency, reduce the labor cost and error rate; through the summation analysis of multi-modal data, the flexibility and adaptability of the check are enhanced, and complex irregular problems can be effectively handled.

[0105] This embodiment addresses scenarios requiring network adjustments. It utilizes neural networks to comprehensively analyze relevant elements such as user call type, call volume, traffic distribution, ARPU, terminal type, terminal functional status, SIM card type, primary / secondary SIM card type, service details, resident cells, and MDT maps to generate feature representations. Simultaneously, a multimodal deep learning model is constructed for feature extraction and fusion, outputting clustered user and network information. Different upgrade strategies are formulated based on different clustering results, ultimately outputting user clustering guidance measures and network optimization and upgrade strategies to assist in the network decommissioning and upgrade process.

[0106] The technical solution of this embodiment involves acquiring user datasets and network datasets; inputting the user datasets and network datasets into a multimodal deep learning model based on a Restricted Boltzmann Machine (RBM); clustering the feature vectors output by the hidden layer of the RBM-based multimodal deep learning model to obtain a clustered profile of users and networks; determining user-residence factors in the first network based on the user-residence profile, wherein the user-residence factors in the first network include at least one of: second network coverage factors, terminal adaptation factors, user intent factors, and value-added service factors; and determining a processing strategy based on the user-residence factors in the first network, wherein the processing strategy includes: expanding the coverage of the second network and generating reminder information. The processing strategy, determined based on the user datasets and network datasets, aims to ensure that users reside in the second network, i.e., the efficient network, thereby improving the utilization rate of wireless resources in the area to be frequency-reduced and decommissioned while guaranteeing the user experience within the area, making the frequency reduction and decommissioning work more orderly and reasonable.

[0107] Example 2

[0108] Figure 5 This is a schematic diagram of a data processing device provided in an embodiment of the present invention. This embodiment is applicable to data processing applications. The device can be implemented using software and / or hardware, and can be integrated into any device that provides data processing functionality, such as… Figure 5 As shown, the data processing device specifically includes: an acquisition module 510, a user and network clustering profile determination module 520, a user residency first network factor determination module 530, and a processing strategy determination module 540.

[0109] The acquisition module is used to acquire user datasets and network datasets.

[0110] The user and network clustering profile determination module is used to input the user dataset and the network dataset into a multimodal deep learning model based on Restricted Boltzmann Machine (RBM), and to cluster the feature vectors output by the hidden layer of the RBM-based multimodal deep learning model to obtain a user and network clustering profile.

[0111] a user staying first network factor determination module configured to determine a user staying first network factor according to the cluster portrait of the user and the network, wherein the user staying first network factor comprises at least one of a second network coverage factor, a terminal adaptation factor, a user intention factor and a value-added service factor;

[0112] a processing strategy determination module configured to determine a processing strategy according to the user staying first network factor, wherein the processing strategy comprises expanding the second network coverage range and generating reminding information.

[0113] The product can execute the method provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of the execution method.

[0114] Embodiment Three

[0115] Figure 6 A structural schematic diagram of an electronic device 10 that can be used to implement embodiments of the application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the applications described and / or claimed in this document.

[0116] As shown in Figure 6 The electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is in communication with the at least one processor 11, wherein the memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0117] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0118] The processor 11 can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the data processing method.

[0119] In some embodiments, the data processing method can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded onto the RAM 13 and executed by the processor 11, one or more steps of the data processing method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the data processing method by any other appropriate means, such as by means of firmware.

[0120] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0121] Computer programs for implementing the methods of the present application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer program, when executed, can cause instructions defined in the flow charts and / or block diagrams to be implemented. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package and partially on a remote machine or entirely on a remote machine or server.

[0122] In the context of the present application, a computer readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, a computer readable storage medium can be a machine readable signal medium. More specific examples of a machine readable storage medium will include one or more lines of a program of instructions in a transitory signal form, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0123] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0124] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), blockchain networks, and the Internet.

[0125] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, and solves the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0126] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the scope of the present application. For example, the steps recited in the present application can be executed in parallel, in series, or in a different order, and the present application is not limited herein as long as the desired results of the technical solutions of the present application can be achieved.

[0127] The embodiments of the present application further provide a computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the data processing method according to any one of the embodiments of the present application.

[0128] Computer program products can be written in any one of a number of programming languages, including object oriented languages such as Java, Smalltalk, C++, or conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0129] The specific embodiments have been shown and described for purposes of illustrating the present application. It will be readily apparent to those skilled in the art that various modifications, combinations, sub-combinations, and alternatives can be made to the specific embodiments without departing from the spirit and principles of the present application. Accordingly, the present application is not intended to be limited to the specific embodiments described herein.

Claims

1. A data processing method, characterized in that, include: Obtain user datasets and network datasets; The user dataset and the network dataset are input into a multimodal deep learning model based on a Restricted Boltzmann Machine (RBM). The feature vectors output by the hidden layer of the RBM-based multimodal deep learning model are clustered to obtain a clustered profile of the user and the network. Based on the clustering profile of users and networks, the first network factor for user residency is determined, wherein the first network factor for user residency includes at least one of the following: second network coverage factor, terminal adaptation factor, user intent factor, and value-added service factor; The processing strategy is determined based on the user's residency in the first network factor, wherein the processing strategy includes: expanding the coverage of the second network and generating reminder information.

2. The method according to claim 1, characterized in that, The user dataset includes at least one of the following: user call type, call volume, network usage frequency, terminal type, average user income, service details data, and complaint data. The network dataset includes at least one of the following: base station operating parameters, resident user type of access base station, number of users, measurement report, network coverage rate of cellular cells, network coverage quality of cellular cells, and minimized roadside MDT network coverage map.

3. The method according to claim 1, characterized in that, Clustering the feature vectors output from the hidden layer of the RBM-based multimodal deep learning model yields a clustered profile of the user and the network, including: The feature vectors output from the hidden layer of the RBM-based multimodal deep learning model are dimensionally compressed to obtain dimensionally compressed feature vectors. Based on the distribution characteristics of the feature vectors after dimensional compression, the target clustering algorithm is determined; Based on the target clustering algorithm, the feature vectors after dimensional compression are clustered to obtain a clustered profile of the user and the network.

4. The method according to claim 1, characterized in that, The RBM-based multimodal deep learning model includes a visible layer and a hidden layer. Both the visible layer and the hidden layer include multiple RBM sub-networks. Each RBM sub-network of the visible layer is connected to all RBM sub-networks of the hidden layer, and each RBM sub-network of the hidden layer is connected to all RBM sub-networks of the visible layer.

5. The method according to claim 4, characterized in that, The training process of the RBM-based multimodal deep learning model includes: Based on the first formula, the activation probability of each RBM subnetwork is determined. The first formula is: in, Let be the activation probability of the i-th RBM subnetwork in the (L+1)-th layer. Let J be the connection weight between the j-th RBM subnetwork in layer L and the ith RBM subnetwork in layer L+1. This represents the state of the j-th RBM subnetwork in layer L; Extract the single-modal features based on the activation probability of each RBM sub-network; Determine the joint probability based on the characteristics of each single mode; Based on the joint probability, unsupervised training is performed on the model to be trained to obtain a multimodal deep learning model based on RBM.

6. The method according to claim 1, characterized in that, Based on the features of each single modality, the joint probability is determined, including: The single-modal features are aggregated to obtain joint abstract features; Based on the second formula, the joint probability is determined. The second formula is: Where v includes the user dataset and the network dataset, h is the joint abstract feature, Θ is the joint probability, P(v,h;Θ) is the joint probability of v and h, Z(Θ) is the partition function, and ε(v,h;Θ) is the energy function.

7. A data processing apparatus, characterized in that, include: The acquisition module is used to acquire user datasets and network datasets. The user and network clustering profile determination module is used to input the user dataset and the network dataset into a multimodal deep learning model based on Restricted Boltzmann Machine (RBM), and to cluster the feature vectors output by the hidden layer of the RBM-based multimodal deep learning model to obtain a user and network clustering profile. The user-residence first network factor determination module is used to determine the user-residence first network factor based on the clustering profile of the user and the network, wherein the user-residence first network factor includes at least one of the following: second network coverage factor, terminal adaptation factor, user intent factor, and value-added service factor; The processing strategy determination module is used to determine a processing strategy based on the user's residency in the first network factor, wherein the processing strategy includes: expanding the coverage of the second network and generating reminder information.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the data processing method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the data processing method according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the data processing method according to any one of claims 1-6.

Images