Network derogatory user prediction analysis method and system, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a network derogatory user prediction analysis method and system, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set, preprocessing the first data set to obtain a second data set, and performing clustering analysis on data in the second data set to obtain a plurality of data clusters; counting a first quantity of data belonging to network derogatory user data and a second quantity of data belonging to non-network derogatory user data in each data cluster, judging whether the data clusters have difficult sample data by comparing the difference between the first quantity and the second quantity of each data cluster, and if the difficult sample data exist, judging whether the data clusters have the difficult sample data; and if not, removing the difficult sample data from the second data set to obtain a third data set to reduce the influence of noise data in the training process, and training the initialized derogatory prediction model by using the third data set to obtain an accurate derogatory prediction model.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, system, electronic device and storage medium for predicting and analyzing network derogatory users. Background Art

[0002] Network depreciation users refer to the user group that has no intention to continue using network resources. With the widespread application of 5G technology, new services are emerging in an endless stream, and users' expectations and requirements for network quality are also constantly increasing. Therefore, how to accurately predict and analyze network depreciation users is of vital importance for operators to improve service quality, optimize network performance, and enhance market competitiveness. In related technologies, depreciation user prediction methods have certain problems in data processing. Although data cleaning, padding, and normalization are performed on users' network usage-related data, the business logic of the user data itself is not deeply considered, resulting in insufficient precision in data processing. This processing method may not be able to reasonably eliminate outliers and noise in the data, thereby affecting the accuracy of the prediction results. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a network derogatory user prediction analysis method, system, electronic device and storage medium, aiming to improve the accuracy of derogatory user prediction results.

[0004] To achieve the above purpose, an embodiment of the present application provides a method for predicting and analyzing network derogatory users, comprising the following steps:

[0005] Acquire network service data of the target user, wherein the network service data includes feature data of different dimensions;

[0006] Inputting the network service data into a derogatory prediction model to obtain a prediction result of the target user, wherein the prediction result is used to determine whether the target user is a network derogatory user;

[0007] The depreciation prediction model is obtained by the following steps:

[0008] Collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set;

[0009] Preprocessing the first data set to obtain a second data set;

[0010] Performing cluster analysis on the network service data in the second data set to obtain multiple data clusters;

[0011] Counting a first quantity of network-derogatory user data and a second quantity of non-network-derogatory user data in each of the data clusters;

[0012] By comparing the difference between the first quantity and the second quantity of each data cluster, determining whether there is difficult sample data in the data cluster;

[0013] When there are difficult sample data in the data cluster, the difficult sample data is removed from the second data set to obtain a third data set;

[0014] The initialized depreciation prediction model is trained using the third data set to obtain a trained depreciation prediction model.

[0015] In some embodiments, collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set includes the following steps:

[0016] Collecting network service satisfaction survey data of user groups, and dividing the user groups into network derogatory user groups and non-network derogatory user groups according to the network service satisfaction survey data;

[0017] Data is screened and extracted in the database according to the identity information of each user in the network derogatory user group and the non-network derogatory user group to obtain network service data of each user to form a first data set.

[0018] In some embodiments, the preprocessing of the first data set to obtain the second data set comprises the following steps:

[0019] Extracting dimensional feature data from the network service data of each user in the first data set respectively to obtain a dimensional feature data set;

[0020] When the amount of data in the dimensional feature data set is less than a first threshold, the dimensional feature data is deleted from the first data set to obtain a second data set.

[0021] In some embodiments, the preprocessing of the first data set to obtain the second data set comprises the following steps:

[0022] Determine the outlier data points of each feature dimension in the first data set by using the median absolute difference method;

[0023] Processing the outlier data points in the first data set according to the outlier data point processing strategy to obtain a second data set;

[0024] Wherein, the outlier data point processing strategy is:

[0025] When the outlier data point is a numerical outlier point, deleting the outlier data point from the first data set;

[0026] When the outlier data point is a logical outlier point, normalization processing is performed on the outlier data point.

[0027] In some embodiments, comparing the difference between the first quantity and the second quantity of each data cluster to determine whether the data cluster contains difficult sample data includes the following steps:

[0028] Multiplying the larger value of the first number and the second number by a preset ratio to obtain a comparison threshold;

[0029] Compare the smaller value of the first quantity and the second quantity with the comparison threshold; when the smaller value is less than the comparison threshold, difficult sample data exists in the data cluster, and the user group data corresponding to the smaller value is difficult sample data.

[0030] In some embodiments, the network derogatory user prediction and analysis method further includes the following steps:

[0031] Determining whether the target user is a network derogatory user according to the prediction result;

[0032] When the target user is a network derogatory user, the network service data of the target user is input into a cause analysis model to obtain the derogatory reasons of the network derogatory user.

[0033] In some embodiments, the cause analysis model is obtained by the following steps:

[0034] Screening and extracting data in a database according to the identity information of each user in the network derogatory user group, obtaining network service data of the network derogatory users, and forming a fourth data set;

[0035] Marking the network service data of each network derogatory user in the fourth data set with a derogatory reason to obtain a fifth data set;

[0036] The initialized cause analysis model is trained using the fifth data set to obtain a trained cause analysis model.

[0037] To achieve the above purpose, another aspect of the embodiment of the present application provides a network derogatory user prediction and analysis system, including:

[0038] The first module is used to obtain network service data of the target user, wherein the network service data includes feature data of different dimensions;

[0039] The second module is used to input the network service data into a derogatory prediction model to obtain a prediction result of the target user, wherein the prediction result is used to determine whether the target user is a network derogatory user;

[0040] The derogation prediction model in the second module is obtained through the first training module, and the first training module is used to:

[0041] Collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set;

[0042] Preprocessing the first data set to obtain a second data set;

[0043] Performing cluster analysis on the network service data in the second data set to obtain multiple data clusters;

[0044] Counting a first quantity of network-derogatory user data and a second quantity of non-network-derogatory user data in each of the data clusters;

[0045] By comparing the difference between the first quantity and the second quantity of each data cluster, determining whether there is difficult sample data in the data cluster;

[0046] When there are difficult sample data in the data cluster, the difficult sample data is removed from the second data set to obtain a third data set;

[0047] The initialized depreciation prediction model is trained using the third data set to obtain a trained depreciation prediction model.

[0048] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application proposes an electronic device, which includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, and the program implements the method described in the above embodiment when executed by the processor.

[0049] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium used for computer-readable storage, and the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described in the above embodiment.

[0050] The network derogatory user prediction and analysis method, system, electronic device and storage medium proposed in the present application collect network service data of network derogatory user groups and non-network derogatory user groups to obtain a first data set, pre-process the first data set to obtain a second data set, and then perform cluster analysis on the network service data in the second data set to obtain multiple data clusters, count the first number of network derogatory user data and the second number of non-network derogatory user data in each data cluster, and judge whether there is difficult sample data in the data cluster by comparing the difference between the first number and the second number of each data cluster. When there is difficult sample data in the data cluster, the difficult sample data is removed from the second data set to obtain a third data set, and then the initialized derogatory prediction model is trained using the third data set to obtain a trained derogatory prediction model. This embodiment traverses all data clusters, deletes the difficult sample data of abnormal and unpredictable users, thereby effectively eliminating noise data caused by abnormal scoring logic, improving the accuracy of training data, and thus improving the accuracy of prediction results of the derogatory prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of a network derogatory user prediction and analysis method provided by an embodiment of the present application;

[0052] Figure 2 yes Figure 1 Flow chart of the derogatory prediction model training method of step S102 in FIG.

[0053] Figure 3 yes Figure 2 Flow chart of step S201 in FIG.

[0054] Figure 4 yes Figure 2 Flow chart of step S202 in FIG.

[0055] Figure 5 yes Figure 2 Another flow chart of step S202 in FIG. 1 ;

[0056] Figure 6 yes Figure 2 Flow chart of step S205 in FIG.

[0057] Figure 7 is a flow chart of a network derogatory user prediction and analysis method provided by another embodiment of the present application;

[0058] Figure 8 yes Figure 7 Flow chart of the cause analysis model training method of step S702;

[0059] Fig. 9 is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application;

[0060] Fig.10 It is a schematic diagram of the overall flow of the network derogatory user prediction and analysis method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] It should be noted that, although the functional modules are divided in the system and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0064] First, some nouns involved in this application are analyzed:

[0065] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0066] MAD (Median Absolute Deviation) is a method for detecting outliers by first calculating the sum of the distances between all factors and the mean.

[0067] The embodiments of the present application provide a network derogatory user prediction and analysis method, system, electronic device and storage medium, aiming to improve the accuracy of derogatory user prediction results.

[0068] The network derogatory user prediction and analysis method, system, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the network derogatory user prediction and analysis method in the embodiments of the present application is described.

[0069] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0070] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0071] The network derogatory user prediction and analysis method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The network derogatory user prediction and analysis method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the network derogatory user prediction and analysis method, etc., but is not limited to the above forms.

[0072] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0073] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0074] Figure 1 is an optional flowchart of the network derogatory user prediction and analysis method provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S102.

[0075] Step S101, obtaining network service data of a target user, where the network service data includes feature data of different dimensions;

[0076] Step S102: input network service data into a derogatory prediction model to obtain a prediction result of a target user, wherein the prediction result is used to determine whether the target user is a network derogatory user.

[0077] In step S101 of some embodiments, network service data refers to various data of users recorded on the network operator platform, and the network service data includes characteristic data of different dimensions, for example, traffic data, user information (gender, age, etc.), complaint data, voice data, location data, etc. The network service data can be multi-domain integrated data, that is, collected from locations such as the BSS domain and the OSS domain.

[0078] In step S102 of some embodiments, the network service data of the target user is input into the derogatory prediction model to determine whether the target user is a network derogatory user. The derogatory prediction model is a classification model that can determine whether the target user is a network derogatory user by calculating the probability that the user's network service data belongs to the network derogatory user label and judging by a threshold. The derogatory prediction model can be a machine learning model such as a support vector machine and a random forest, or a neural network model based on deep learning.

[0079] Please refer to Figure 2 The depreciation prediction model in step 102 can be obtained by, but not limited to, the following steps:

[0080] Step S201, collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set;

[0081] Step S202, preprocessing the first data set to obtain a second data set;

[0082] Step S203, performing cluster analysis on the network service data in the second data set to obtain multiple data clusters;

[0083] Step S204, counting a first quantity of network derogatory user data and a second quantity of non-network derogatory user data in each data cluster;

[0084] Step S205, by comparing the difference between the first quantity and the second quantity of each data cluster, determining whether there is difficult sample data in the data cluster;

[0085] Step S206, when there are difficult sample data in the data cluster, the difficult sample data is removed from the second data set to obtain a third data set;

[0086] Step S207: train the initialized depreciation prediction model using the third data set to obtain a trained depreciation prediction model.

[0087] In step S201 of some embodiments, the first data set includes network service data of a plurality of users, and the first data set divides the users into two categories: network derogatory users and non-network derogatory users.

[0088] In step S202 of some embodiments, the preprocessing operation of the first data set includes but is not limited to cleaning obvious outliers, filling missing data, data normalization, feature mapping and other preprocessing operations, and the original data is converted into training data that can be processed by a computer. The feature mapping preprocessing operation refers to mapping non-numeric character type feature data to discrete integer type data, such as when the user's "gender" is "male" or "female", its feature value is mapped to "[1,0]" or "[0,1]" respectively.

[0089] In step S203 of some embodiments, a clustering algorithm may be used to perform cluster analysis on the network service data in the second data set, and the clustering algorithm may be a K-Means algorithm, a K-Medoids algorithm, a DBSCAN algorithm, etc. Exemplarily, this embodiment may use a K-means method, which uses distance to describe the similarity between two data objects.

[0090] The k-means algorithm is a partition-based clustering algorithm. The algorithm uses k as a parameter to divide n data objects into k clusters, so that the similarity within the cluster is high, while the similarity between clusters is low. The processing flow of the k-means algorithm is as follows. First, randomly select k data objects, each of which represents a cluster center, that is, select k initial centers; for each remaining object, assign it to the cluster corresponding to the cluster center that is most similar to it according to its distance from each cluster center; then recalculate the average value of all objects in each cluster as the new cluster center. Repeat the above process until the criterion function converges, that is, the cluster center does not change significantly. The mean square error is usually used as the criterion function, that is, minimize the sum of the squares of the distance from each point to the nearest cluster center.

[0091] The k-means algorithm clustering steps are as follows:

[0092] S1. Randomly select k user data objects [x 1 ,x 2 ,…,x k ], each user data represents a cluster center in the high-dimensional feature space;

[0093] S2, traverse the remaining user data in the data set, and use the Euclidean space 2 norm as a metric to calculate the user feature vector x and the center of each cluster [x 1 ,x 2 ,…,x k ]'s 2-norm || xx k || 2 , and finally assign each user data to the cluster with the smallest 2-norm;

[0094] S3. Recalculate the average of all user feature vectors in each cluster and use the average feature vector as the new cluster center.

[0095] S4. Repeat steps 2 and 3 above until the mean square error change of each cluster is less than the set threshold. At this time, the K-means algorithm is considered to have converged, and the user data is divided into k categories, that is, k data clusters are obtained.

[0096] In step S204 of some embodiments, for the k data clusters obtained by the clustering algorithm, each data cluster contains network service data of several users. Through the network derogatory user labels and non-network derogatory user labels of the network service data, the first number of network derogatory user data in the data cluster (i.e., network service data belonging to network derogatory users) and the second number of non-network derogatory user data in the data cluster (i.e., network service data belonging to non-network derogatory users) can be counted.

[0097] In step S205 of some embodiments, by comparing the gap between the first quantity and the second quantity in each data cluster, it can be determined whether there is difficult sample data in the data cluster, wherein the difficult sample data refers to the network service data of unpredictable users. If the difference in the quantity of two different types of data in the data cluster is too large, it means that the data corresponding to the minority category labels is abnormal and unpredictable, and this part of the data is considered to be difficult sample data. The gap analysis method can be a method of directly comparing the first quantity and the second quantity, or a method of analyzing the proportion of the first quantity and the second quantity.

[0098] In step S206 of some embodiments, after the difficult sample data is obtained through inter-class analysis, it is necessary to perform invalid filtering on these data, that is, traverse all data clusters, delete the network business data of users that are judged to be abnormal and unpredictable in the second data set to obtain a third data set, thereby effectively eliminating noise data caused by abnormal scoring logic and improving the accuracy of training data.

[0099] In step S207 of some embodiments, the initialized depreciation prediction model is trained using the third data set to obtain a trained depreciation prediction model.

[0100] Exemplarily, the prediction methods in the related art mainly rely on simple models such as logistic regression and random forests. These models often fail to achieve satisfactory prediction results when faced with massive data. In particular, when dealing with complex and changeable user data, the generalization ability and stability of these models may be seriously challenged. In order to accurately predict derogatory users, this embodiment uses a multi-granularity cascade forest method to construct a non-neural network deep model, namely a deep forest. The deep forest has a cascade structure, and feature learning is performed by enhancing the cascade forest through multi-granularity scanning. The network derogatory user prediction problem is regarded as a classification problem, and the user sample data is divided into two categories according to the degree of satisfaction with the wireless network: satisfied and dissatisfied. Among them, the wireless network satisfaction score is higher than 7 points for satisfied users, and the satisfaction score is 7 points or less for dissatisfied users. The deep forest model is trained to predict network derogatory users.

[0101] Each layer of the deep forest cascade structure contains several ensemble learning classifiers, which is an integrated structure in an integration. The deep forest built in this embodiment uses two basic ensemble learning classifiers, namely, a completely random forest and an ordinary random forest. A completely random forest randomly selects features to split in the complete feature space. The ordinary random forest selects split nodes by the Gini coefficient in a random feature subspace. This embodiment uses two completely random forests and two random forests, each of which contains 500 completely random trees, and the number of trees in the forest is a hyperparameter. After inputting the network service data into the random forest, an estimated class distribution can be obtained, which is connected to the original feature vector to be input into the next cascade cascade. For example, assuming there are 2 classes, each of the four random forests will generate a 2-dimensional class vector, so the next level of cascade will obtain 8 enhanced features.

[0102] The training of each forest is independent supervised learning, and each forest outputs a classification probability distribution, which is equivalent to a new feature. These new features are combined with the original input features to form a new input for the generation of the next layer. After each layer of forest is trained, the forest of that layer is used on the validation set to see if the performance continues to improve. If the improvement is not obvious, no new layers will be generated. After each layer of forest is generated and trained, the forest of that layer is used on the test set to verify if the performance continues to improve. If the improvement is not obvious, no new layers will be generated. In other words, the iteration stops when the effect cannot be improved.

[0103] Multi-granularity scanning is a technical means to do more processing on features. The specific scanning process is similar to the movement of the convolution kernel in the convolutional neural network, but no convolution operation is performed. Like the processing of image data and sequence data, the image data is scanned from left to right and from top to bottom, while the sequence data is scanned from top to bottom. This operation mainly expands the dimension of the data and pays more attention to the relationship between the various features in the original data.

[0104] In order to improve the processing capability when facing massive data, this embodiment builds a deep forest network model consisting of a cascade forest structure and multi-granularity scanning, and uses objective data to train the model. When facing new data streams, it can efficiently and accurately predict potential derogatory users.

[0105] This embodiment utilizes the category label distribution characteristics to effectively discover a small number of difficult sample data in the data cluster. By using the third data set after eliminating the difficult sample data to train the derogation prediction model, the impact of noise data on the derogation prediction model can be reduced, thereby improving the model accuracy.

[0106] See also Figure 3 In some embodiments, step S201 may include but is not limited to steps S301 to S302:

[0107] Step S301, collecting network service satisfaction survey data of user groups, and dividing the user groups into network derogatory user groups and non-network derogatory user groups according to the network service satisfaction survey data;

[0108] Step S302, data is screened and extracted from a database according to the identity information of each user in the network derogatory user group and the non-network derogatory user group to obtain network service data of each user, and form a first data set.

[0109] In this embodiment, data can be collected from BSS domains, OSS domains and other locations to extract user satisfaction perception survey data, wherein the survey data can be obtained by calling the operator's customer service outbound survey system to obtain customer satisfaction score scores, and then the user groups are divided according to the scores. For example, taking a 10-point satisfaction score as an example, users with satisfaction scores of 1-7 points are divided into network derogatory users, and the derogatory attribute of the derogatory users is set to 1, and the derogatory attribute of the non-network derogatory users is set to 0. Next, the user product instance ID is aligned for association screening, and data is screened in a multi-domain database by user ID, and multiple data are merged into the user's network service data, and the first data set is obtained based on the network service data of multiple users.

[0110] See also Figure 4 In some embodiments, step S202 may include but is not limited to steps S401 to S402:

[0111] Step S401, extracting dimensional feature data from the network service data of each user in the first data set to obtain a dimensional feature data set;

[0112] Step S402: when the amount of data in the dimensional feature data set is less than a first threshold, the dimensional feature data is deleted from the first data set to obtain a second data set.

[0113] In this embodiment, the preprocessing process includes cleaning the dimensional data with more missing data in the first data set. Specifically, the network service data of each user in the first data set, each network service data contains feature data of different dimensions, and a certain dimensional feature data is extracted from the network service data of each user to obtain the dimensional feature data set. If the amount of data in the dimensional feature data set is less than the first threshold, it means that there are too many missing data in this dimension, then the dimensional feature data is deleted from the first data set to obtain the second data set, thereby improving the data availability for training the model. Among them, the first threshold can be obtained by multiplying the total amount of network service data in the first data set by a preset ratio. Furthermore, for the remaining feature dimensions, there are still missing data therein, and the samples containing missing data are deleted from the data set.

[0114] See also Figure 5 In some embodiments, step S202 may also include but is not limited to steps S501 to S502:

[0115] Step S501, using the median absolute difference method to determine the outlier data points of each feature dimension in the first data set;

[0116] Step S502, processing the outlier data points in the first data set according to the outlier data point processing strategy to obtain a second data set;

[0117] Among them, the outlier data point processing strategy is:

[0118] When the outlier data point is a numerical outlier, the outlier data point is deleted from the first data set;

[0119] When the outlier data point is a logical outlier, the outlier data point is normalized.

[0120] In this embodiment, the preprocessing process also includes processing the outlier data points in the first data set. Specifically, the Median Absolute Deviation (MAD) method is first used to find the outlier data points. MAD is a method that first calculates the sum of the distances between all factors and the mean to detect outliers. The first step of the processing process is to find the median Xmedian of all factors; the second step is to obtain the absolute deviation value Xi-Xmedian of each factor from the median; the third step is to obtain the median M of the absolute deviation value; finally, the parameter n is determined, so that the reasonable range is determined to be [Xmedian-nM, Xmedian+nM]. For those that exceed the reasonable range, they are determined to be outlier data points.

[0121] For the outlier data points found, consider their business meaning and distinguish between numerical outliers and logical outliers. Numerical outliers are those whose feature attribute values ​​obviously exceed the defined range. For example, the value of "user call quality score" is 10 points higher than the highest score of the questionnaire. For such data, delete them directly. Logical outliers are those whose feature attributes have emergencies. The abnormality of the data comes from the scarcity of the user's behavior. Directly deleting the data will lead to the loss of important information. Therefore, retain the data and eliminate the abnormality through normalization in subsequent operations.

[0122] The data normalization process is as follows:

[0123] All features of the data sample have the same influence on the sample distance in the feature space. However, after the features are digitized, due to the different values, the distance between sample points in the space will be dominated by individual feature values ​​and will be less affected by other features. Therefore, all data needs to be mapped to the same scale. This embodiment uses the MinMax normalization method to perform linear transformation on the data and map the feature values ​​to [0, 1]. The conversion function is as follows:

[0124]

[0125] Among them, x max is the maximum value of the characteristic attribute, x min is the minimum value of the characteristic attribute.

[0126] See also Figure 6 In some embodiments, step S205 may include but is not limited to steps S601 to S602:

[0127] Step S601, multiplying the larger value of the first number and the second number by a preset ratio to obtain a comparison threshold;

[0128] Step S602: compare the smaller quantity value of the first quantity and the second quantity with a comparison threshold. When the smaller quantity value is less than the comparison threshold, difficult sample data exists in the data cluster, and the user group data corresponding to the smaller quantity value is difficult sample data.

[0129] In this embodiment, for the data clusters obtained by clustering algorithm, this embodiment adopts the method of inter-class analysis and filtering to judge and eliminate the network service data of unpredictable users. The specific method of inter-class analysis is: for each data cluster, the number of different category labels (network derogatory users and non-network derogatory users) in the data cluster is counted. If the number of a certain category label is less than the preset ratio (such as 50%) of the number of another category label, it is considered that the data corresponding to the minority category label is unpredictable, that is, difficult sample data. For example, in a certain data cluster, there are 80 label data of non-network derogatory users and 20 labels of network derogatory users. The number of labels of network derogatory users (20) is less than 50% of the number of labels of non-network derogatory users (80), so the 20 data corresponding to network derogatory users in the data cluster are considered abnormal and unpredictable. After the difficult sample data is obtained by inter-class analysis, it is necessary to perform invalid filtering on these data, that is, traverse all data clusters and delete the user data judged as abnormal and unpredictable according to the rule, so as to effectively eliminate the noise data caused by the abnormal scoring logic and improve the accuracy of the training data.

[0130] See also Figure 7 In some embodiments, the network derogatory user prediction and analysis method of the embodiment of the present application may also include but is not limited to steps S701 to S702:

[0131] Step S701, judging whether the target user is a network derogatory user according to the prediction result;

[0132] Step S702: When the target user is a network derogatory user, the network service data of the target user is input into a cause analysis model to obtain the derogatory cause of the network derogatory user.

[0133] In this embodiment, the prediction methods in the related art often lack in-depth analysis of the reasons for network derogatory users. Derogatory users may choose to leave for a variety of reasons, including but not limited to insufficient network coverage, unstable signals, unreasonable package prices, poor customer service, etc. It is difficult to help operators improve service quality by simply predicting derogatory users. Based on this, this embodiment inputs the network service data of network derogatory users into a cause analysis model on the basis of predicted potential derogatory users to obtain the derogatory reasons of network derogatory users. The cause analysis model can determine the user's derogatory reasons by calculating the probability that the user's network service data belongs to the derogatory reason label, thereby improving service quality and user experience in a targeted manner. The cause analysis model can be a machine learning model such as a support vector machine, a random forest, a deep forest, or a neural network model based on deep learning.

[0134] See also Figure 8In some embodiments, the cause analysis model of step S702 can be obtained through but not limited to steps S801 to S703:

[0135] Step S801, screening and extracting data in a database according to the identity information of each user in the network derogatory user group, obtaining network service data of the network derogatory users, and forming a fourth data set;

[0136] Step S802, marking the derogatory reason for the network service data of each network derogatory user in the fourth data set to obtain a fifth data set;

[0137] Step S803: Use the fifth data set to train the initialized cause analysis model to obtain a trained cause analysis model.

[0138] In this embodiment, a fourth data set can be formed for the network service data of the network derogatory users. The network service data in the fourth data set can be collected from the database, and the network service data of the network derogatory users can be obtained by extracting the network service data from the second data set. For the data of the derogatory users included in the fourth data set, the derogatory reason label is marked by manual labeling to obtain the fifth data set. The users are divided into four categories according to the derogatory reasons, namely, Internet derogation, call derogation, fee derogation and service derogation. Therefore, the cause analysis problem is regarded as a 4-classification problem, and the above-mentioned deep forest can also be used for classification. Let the output feature of each random decision tree classifier in the deep forest be 4, indicating that the number of classification targets of the deep forest is 4, and multi-granularity scanning is used to further increase the ability of the deep forest to capture the correlation of local features. Perform separate supervised learning on each random forest in the deep forest, so that the random forest can output new features. If the feature classification result output by the random forest of the current layer is close to the label, stop generating a new forest, and the cause analysis model training is completed. The data of the predicted network derogatory users is input into the model to analyze the derogatory reasons, so as to optimize the service in a targeted manner and improve user satisfaction.

[0139] According to some embodiments of the present application, the overall process of the network derogatory user prediction and analysis method of the embodiments of the present application is described.

[0140] First, user data is selected from each domain and network behavior attribute database through big data to obtain multi-domain integrated data. The dimensional feature data of each domain are as follows:

[0141] Table 1 BSS domain data

[0142]

[0143]

[0144] Table 2 OSS domain data

[0145]

[0146] Table 3 User survey data

[0147]

[0148]

[0149] The user ID is used to filter data in the multi-domain database, and the feature data of different dimensions are merged to obtain the network business data of each user.

[0150] By cleaning the original data, the data with missing features are removed. Then the MAD method is used to judge the outlier data points. For those that exceed the reasonable range, they are determined to be numerical outliers and deleted. The logical outliers are normalized. After data cleaning, outlier analysis and feature mapping, the second data set with reduced data is obtained.

[0151] The preprocessed data is clustered using the K-means algorithm to calculate the relationship between each sample point x and each cluster center [x 1 ,x 2 ,…,x k ]'s second norm || xx k || 2 Finally, each user data is assigned to the cluster with the smallest two-norm. The difficult sample data is determined by the difference in the number of data with different category labels in the data cluster, and then the difficult sample data is invalidated to obtain the third data set.

[0152] The above data are used to build a derogation prediction model and a cause analysis model. Both the derogation prediction model and the cause analysis model can adopt a deep forest cascade structure, and use multi-granularity scanning technology to analyze the data features. The derogation user data set input is the total feature of the data, which is marked with 2 categories: satisfied and dissatisfied. The wireless network satisfaction score is higher than 7 points for satisfied users, and the satisfaction score is 7 points or less for dissatisfied users. The cause analysis data set input is the total feature of the data, which is marked with 4 categories: Internet derogation, call derogation, tariff derogation and service derogation. Then two deep forests are constructed using the multi-granularity cascade forest method. The derogation user prediction deep forest contains four random forests, each of which contains 500 completely random trees, and the number of output classifications is 2. The cause analysis deep forest also contains four random forests, each of which contains 500 completely random trees, and the number of output classifications is 4. The training process is as described above and will not be repeated here.

[0153] After the model is trained, the user's network service data is input into the derogation prediction model to obtain the result of whether the user is a network derogatory user. Then the data of the network derogatory user is input into the cause analysis model to determine the reason for derogation. For example, in this prediction and analysis process, the model found 643 derogatory users among 29,756 valid users. Among the derogatory reasons of derogatory users, "dissatisfaction with signal quality" accounted for the highest proportion, accounting for 43%. Therefore, in the next stage, it is necessary to deploy more base stations, adjust and optimize the antenna layout, and improve the signal quality.

[0154] The embodiment of the present application also provides a network derogatory user prediction and analysis system, including:

[0155] The first module is used to obtain the network service data of the target user, and the network service data includes feature data of different dimensions;

[0156] The second module is used to input the network service data into the derogatory prediction model to obtain the prediction result of the target user, wherein the prediction result is used to determine whether the target user is a network derogatory user;

[0157] The depreciation prediction model in the second module is obtained through the first training module, and the first training module is used to:

[0158] Collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set;

[0159] Preprocessing the first data set to obtain a second data set;

[0160] Performing cluster analysis on the network service data in the second data set to obtain multiple data clusters;

[0161] Counting a first quantity of network-derogatory user data and a second quantity of non-network-derogatory user data in each data cluster;

[0162] By comparing the difference between the first quantity and the second quantity of each data cluster, determining whether there is difficult sample data in the data cluster;

[0163] When there are difficult sample data in the data cluster, the difficult sample data is removed from the second data set to obtain a third data set;

[0164] The initialized depreciation prediction model is trained using the third data set to obtain a trained depreciation prediction model.

[0165] It can be understood that the contents of the above-mentioned network derogatory user prediction and analysis method embodiments are applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those in the above-mentioned network derogatory user prediction and analysis method embodiments, and the beneficial effects achieved are also the same as those achieved by the above-mentioned network derogatory user prediction and analysis method embodiments.

[0166] The embodiment of the present application also provides an electronic device, the electronic device comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, and the program is executed by the processor to realize the above-mentioned network derogatory user prediction and analysis method. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0167] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0168] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0169] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the network derogatory user prediction and analysis method of the embodiment of this application;

[0170] Input / output interface 903, used to implement information input and output;

[0171] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0172] A bus 905 that transmits information between the various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0173] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0174] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned network derogatory user prediction and analysis method.

[0175] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0176] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0177] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0178] The system embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0179] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0180] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0181] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0182] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0183] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0184] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0185] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0186] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for predicting and analyzing network derogatory users, characterized in that: The following steps are involved: Acquire network service data of the target user, wherein the network service data includes feature data of different dimensions; Inputting the network service data into a derogatory prediction model to obtain a prediction result of the target user, wherein the prediction result is used to determine whether the target user is a network derogatory user; The depreciation prediction model is obtained by the following steps: Collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set; Preprocessing the first data set to obtain a second data set; Performing cluster analysis on the network service data in the second data set to obtain multiple data clusters; Counting a first quantity of network-derogatory user data and a second quantity of non-network-derogatory user data in each of the data clusters; By comparing the difference between the first quantity and the second quantity of each data cluster, determining whether there is difficult sample data in the data cluster; When the data cluster contains difficult sample data, the difficult sample data is removed from the second data set to obtain a third data set; The initialized depreciation prediction model is trained using the third data set to obtain a trained depreciation prediction model.

2. The network derogatory user prediction and analysis method according to claim 1, characterized in that: The collecting of network service data of the network derogatory user group and the non-network derogatory user group to obtain a first data set comprises the following steps: Collecting network service satisfaction survey data of user groups, and dividing the user groups into network derogatory user groups and non-network derogatory user groups according to the network service satisfaction survey data; Data is screened and extracted in the database according to the identity information of each user in the network derogatory user group and the non-network derogatory user group to obtain network service data of each user to form a first data set.

3. The network derogatory user prediction and analysis method according to claim 1, characterized in that: The preprocessing of the first data set to obtain the second data set comprises the following steps: Extracting dimensional feature data from the network service data of each user in the first data set respectively to obtain a dimensional feature data set; When the amount of data in the dimensional feature data set is less than a first threshold, the dimensional feature data is deleted from the first data set to obtain a second data set.

4. The network derogatory user prediction and analysis method according to claim 1, characterized in that: The preprocessing of the first data set to obtain the second data set comprises the following steps: Determine the outlier data points of each feature dimension in the first data set by using the median absolute difference method; Processing the outlier data points in the first data set according to the outlier data point processing strategy to obtain a second data set; Wherein, the outlier data point processing strategy is: When the outlier data point is a numerical outlier point, deleting the outlier data point from the first data set; When the outlier data point is a logical outlier point, normalization processing is performed on the outlier data point.

5. The network derogatory user prediction and analysis method according to claim 1, characterized in that: The step of comparing the difference between the first quantity and the second quantity of each data cluster to determine whether the data cluster has difficult sample data comprises the following steps: Multiplying the larger value of the first number and the second number by a preset ratio to obtain a comparison threshold; Compare the smaller value of the first quantity and the second quantity with the comparison threshold; when the smaller value is less than the comparison threshold, difficult sample data exists in the data cluster, and the user group data corresponding to the smaller value is difficult sample data.

6. The network derogatory user prediction and analysis method according to claim 1, characterized in that: The network derogatory user prediction and analysis method further comprises the following steps: Determining whether the target user is a network derogatory user according to the prediction result; When the target user is a network derogatory user, the network service data of the target user is input into a cause analysis model to obtain the derogatory reasons of the network derogatory user.

7. The network derogatory user prediction and analysis method according to claim 6, characterized in that: The cause analysis model is obtained by the following steps: Screening and extracting data in a database according to the identity information of each user in the network derogatory user group, obtaining network service data of the network derogatory users, and forming a fourth data set; Marking the network service data of each network derogatory user in the fourth data set with a derogatory reason to obtain a fifth data set; The initialized cause analysis model is trained using the fifth data set to obtain a trained cause analysis model.

8. A network derogatory user prediction and analysis system, characterized in that: include: The first module is used to obtain network service data of the target user, wherein the network service data includes feature data of different dimensions; The second module is used to input the network service data into a derogatory prediction model to obtain a prediction result of the target user, wherein the prediction result is used to determine whether the target user is a network derogatory user; The derogation prediction model in the second module is obtained through the first training module, and the first training module is used to: Collecting network service data of a network derogatory user group and a non-network derogatory user group to obtain a first data set; Preprocessing the first data set to obtain a second data set; Performing cluster analysis on the network service data in the second data set to obtain multiple data clusters; Counting a first quantity of network-derogatory user data and a second quantity of non-network-derogatory user data in each of the data clusters; By comparing the difference between the first quantity and the second quantity of each data cluster, determining whether there is difficult sample data in the data cluster; When the data cluster contains difficult sample data, the difficult sample data is removed from the second data set to obtain a third data set; The initialized depreciation prediction model is trained using the third data set to obtain a trained depreciation prediction model.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the method described in any one of claims 1 to 7 are realized.

10. A storage medium, the storage medium being a computer-readable storage medium, used for computer-readable storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of claims 1 to 7.

Images