Network service platform adjusting method and device, electronic equipment and medium

Abstract

The embodiment of the invention discloses a network service platform adjusting method and device, electronic equipment and a medium. A specific embodiment of the method comprises the following steps: acquiring a user data set and a user comment information set from different sources; performing user anomaly detection on the fused user data set to obtain an abnormal user information set; removing an abnormal user information set from the fused user data set to obtain a target user information set; performing user clustering processing on the target user information set to obtain a user class cluster set; performing user churn rate prediction on the user class cluster set to obtain a user churn rate numerical value set; screening user churn rate values greater than or equal to a preset user churn rate threshold to obtain a target user churn rate value set; performing demand prediction on the user comment information set to obtain a user demand information set; and dynamically adjusting the network service platform. According to the embodiment, the user data processing efficiency can be improved, the user demand can be accurately predicted, and the experience feeling of the user on the network service platform is improved.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technologies, and more particularly, to methods, apparatuses, electronic devices, and media for adjusting a network service platform. Background Art

[0002] User data analysis can establish the association between users and a network service platform. By deeply tracking user data and analyzing user needs, the user experience can be improved. For the adjustment of a network service platform, the commonly used method is as follows: An association rule algorithm is used to perform anomaly detection on a user data set in a data source to obtain an abnormal user data set. A density clustering algorithm is used to cluster the target user data set obtained by removing the abnormal user data set from the user data set to obtain a user clustering cluster set. The user churn rate of each user clustering cluster in the user clustering cluster set is predicted to improve the user retention rate.

[0003] However, the inventors found that when the above method is used to adjust a network service platform, the following technical problems often exist:

[0004] First, due to clustering using the density clustering algorithm and the data source being a single data source, the data quality of user data in a single data source is likely to be low, resulting in a low accuracy of user data analysis. In addition, the initialization of the density clustering algorithm is random and subjective, which is likely to cause a low clustering accuracy and a low user experience.

[0005] Second, when using the association rule algorithm for anomaly detection, since the association rule algorithm needs to perform multiple comprehensive searches on the target user data set, it consumes a large amount of computing resources, resulting in a low anomaly detection efficiency and a long detection time, leading to a low security of user data and an increased system load.

[0006] The above information disclosed in this background art section is only used to enhance the understanding of the background of the inventive concept, and thus, it may include information that does not constitute the prior art known to those of ordinary skill in the art in this country. Summary of the Invention

[0007] This summary of the disclosure is provided to introduce concepts in a brief form that will be described in detail in the subsequent detailed description section. This summary of the disclosure is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to be used to limit the scope of the claimed technical solution.

[0008] Some embodiments of the present disclosure propose methods, apparatuses, electronic devices, and media for adjusting a network service platform to solve one or more of the technical problems mentioned in the above background art section.

[0009] In a first aspect, some embodiments of the present disclosure provide a method for adjusting a network service platform, including: obtaining user data sets and user comment information sets from different sources; performing data fusion on the user data sets from different sources to obtain a fused user data set; performing user anomaly detection on the fused user data set to obtain an abnormal user information set; removing the abnormal user information set from the fused user data set to obtain a user information set after removal, which is used as a target user information set; performing user clustering processing on the target user information set to obtain a user cluster set; predicting the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set; screening out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set; for each target user cluster in the target user cluster set, perform the following adjustment steps: selecting at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set, where the target user cluster set is each user cluster corresponding to the target user churn rate value set; performing demand prediction on the target user comment information set to obtain a user demand information set; dynamically adjusting the network service platform corresponding to the user demand information set according to the user demand information set.

[0010] In a second aspect, some embodiments of the present disclosure provide a network service platform adjustment device, including: an acquisition unit configured to acquire user data sets and user comment information sets from different sources; a data fusion unit configured to perform data fusion on the user data sets from different sources to obtain a fused user data set; a user anomaly detection unit configured to perform user anomaly detection on the fused user data set to obtain an abnormal user information set; a removal unit configured to remove the abnormal user information set from the fused user data set to obtain a removed user information set as a target user information set; a user clustering processing unit configured to perform user clustering processing on the target user information set to obtain a user cluster set; a user churn rate prediction unit configured to predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set; a screening unit configured to screen out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set; an execution unit configured to perform the following adjustment steps for each target user cluster in the target user cluster set: select at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set, where the target user cluster set is each user cluster corresponding to the target user churn rate value set; perform demand prediction on the target user comment information set to obtain a user demand information set; and dynamically adjust the network service platform corresponding to the user demand information set according to the user demand information set.

[0011] In a third aspect, some embodiments of the present disclosure provide an electronic device, including: one or more processors; a storage device storing one or more programs thereon, and when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method described in any implementation manner of the first aspect.

[0012] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium storing a computer program thereon, where the computer program, when executed by a processor, implements the method described in any implementation manner of the first aspect.

[0013] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: The network service platform adjustment method of some embodiments of the present disclosure can improve the efficiency of user data processing and accurately predict user needs, thereby enhancing the user experience on the network service platform. Specifically, the reasons for the relatively low user experience are as follows: Since density clustering algorithm is used for clustering and the data source is a single data source, it is easy for the data quality of user data in a single data source to be low, resulting in a low accuracy rate of user data analysis. In addition, the initialization of the density clustering algorithm is random and subjective, which is likely to lead to a low clustering accuracy rate and thus a low user experience. Based on this, the network service platform adjustment method of some embodiments of the present disclosure can, first, obtain user data sets and user comment information sets from different sources. Here, the integrity and accuracy of the data can be improved, which is used for subsequent user clustering and user demand analysis. Second, perform data fusion on the above-mentioned user data sets from different sources to obtain a fused user data set. Here, data fusion can improve the consistency, integrity, and comprehensiveness of the user data set, and enhance the quality of the user data set. Third, perform user anomaly detection on the above-mentioned fused user data set to obtain an abnormal user information set. Here, user anomaly detection can detect abnormal user behaviors and improve the security of user data. Then, remove the above-mentioned abnormal user information set from the above-mentioned fused user data set to obtain a user information set after removal, which is used as the target user information set. Here, the interference of useless data and the data volume can be reduced, and the system operation load can be alleviated. Subsequently, perform user clustering processing on the above-mentioned target user information set to obtain a user cluster set. Here, performing user clustering processing on the target user information set can identify the user preferences and needs of user clusters, thereby improving the user experience. After that, predict the user churn rate for each user cluster in the above-mentioned user cluster set to obtain a user churn rate value set. Here, predicting the user churn rate can accurately predict the factors that affect the reduction of the user experience, reduce the user churn rate, and enhance the competitiveness and service level of the network service platform. Then, select at least one user churn rate value greater than or equal to a preset user churn rate threshold from the above-mentioned user churn rate value set to obtain a target user churn rate value set. Here, it is convenient for subsequent demand analysis of user clusters with a user churn rate greater than or equal to the preset user churn rate threshold to reduce the risk of user churn and improve the user experience. Finally, for each target user cluster in the target user cluster set, perform the following adjustment steps: First step, select at least one user comment information corresponding to the above-mentioned target user cluster from the above-mentioned user comment information set to obtain a target user comment information set, where the above-mentioned target user cluster set is each user cluster corresponding to the above-mentioned target user churn rate value set. Here, selecting from the user comment information set can improve the accuracy of user comment information, reduce the data volume for subsequent demand prediction, and reduce the system operation load.In the second step, demand prediction is performed on the above-mentioned target user comment information set to obtain a user demand information set. Here, demand prediction can fully understand the needs of users, so as to improve the user experience and user retention rate. In the third step, according to the above-mentioned user demand information set, the network service platform corresponding to the above-mentioned user demand information set is dynamically adjusted. Here, the user experience and the service level of the network service platform can be improved. Thus, it can be seen that this network service platform adjustment method can improve the efficiency of user data processing and accurately predict user needs by performing anomaly detection, clustering, churn rate identification, and demand prediction processing on user data sets from different sources, improve the user experience on the network service platform, and improve the service performance of the network service platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In combination with the accompanying drawings and with reference to the following specific embodiments, the above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent. Throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic and the elements and elements are not necessarily drawn to scale.

[0015] Figure 1 is a flowchart of some embodiments of a network service platform adjustment method according to the present disclosure;

[0016] Figure 2 is a schematic structural diagram of some embodiments of a network service platform adjustment device according to the present disclosure;

[0017] Figure 3 is a schematic structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0019] In addition, it should be noted that for the sake of convenience of description, only parts related to the relevant invention are shown in the drawings. Without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other.

[0020] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules, or units, and are not used to limit the order of functions performed by these devices, modules, or units or their interdependent relationships.

[0021] It should be noted that the modification of "one" and "multiple" mentioned in this disclosure is illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise clearly specified in the context, it should be understood as "one or more".

[0022] The names of the messages or information exchanged between multiple devices in the embodiments of this disclosure are only for illustrative purposes and do not limit the scope of these messages or information.

[0023] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0024] Figure 1 Flow 100 of some embodiments of the network service platform adjustment method according to the present disclosure is shown. The network service platform adjustment method includes the following steps:

[0025] Step 101, obtain user data sets and user comment information sets from different sources.

[0026] In some embodiments, the execution subject (such as an electronic device) of the above network service platform adjustment method can obtain user data sets and user comment information sets from different sources through a wired connection method or a wireless connection method. Among them, the user data in the above user data set can be data related to users. The above user data may include, but is not limited to, at least one of the following: user basic information, user purchase data, user browsing data, and user consultation data. The user comment information in the above user comment information set can be text information published by users for items. The above user data sets from different sources can be user data sets from multiple network service platforms (e-commerce platforms).

[0027] Step 102, perform data fusion on the user data sets from different sources to obtain a fused user data set.

[0028] In some embodiments, the execution subject can perform data fusion on the user data sets from different sources to obtain a fused user data set. Among them, the fused user data in the above fused user data set can be user data after removing duplicates and semantic ambiguities.

[0029] As an example, the above-mentioned execution entity may first use an ETL (Extraction, Transformation, Loading) tool to clean each user dataset in the above-mentioned user datasets from different sources, and obtain a set of cleaned user data groups. Among them, the cleaned user data in the above-mentioned set of cleaned user data groups may be user data obtained after removing missing values, error data, duplicate removal, data mapping, and format conversion. Then, perform data matching on the above-mentioned set of cleaned user data groups to obtain a matching result set. Finally, integrate the set of cleaned user data groups corresponding to at least one matching result indicating successful matching to obtain a fused user dataset.

[0030] Step 103: Perform user anomaly detection on the fused user dataset to obtain a set of abnormal user information.

[0031] In some embodiments, the above-mentioned execution entity may perform user anomaly detection on the above-mentioned fused user dataset to obtain a set of abnormal user information. Among them, the abnormal user information in the above-mentioned set of abnormal user information may be user information with abnormal behavior. In practice, the above-mentioned execution entity may use a user anomaly detection model to perform user anomaly detection on the fused user dataset to obtain a set of abnormal user information. Among them, the above-mentioned user anomaly detection model may be a convolutional neural network model.

[0032] In some optional implementation manners of some embodiments, the above-mentioned performing user anomaly detection on the above-mentioned fused user dataset to obtain a set of abnormal user information may include the following steps:

[0033] First step: Extract features from the user behavior data included in the above-mentioned fused user dataset to obtain a set of user behavior feature vectors. Among them, the above-mentioned user behavior data may be the behavior data of a user collected by a network service platform on the network service platform. For example, the above-mentioned user behavior data may include, but is not limited to, at least one of the following: purchase records of item information, browsing records, and adding to shopping cart records. The user behavior feature vectors in the above-mentioned set of user behavior feature vectors may represent the feature information of the user behavior data.

[0034] In the second step, format conversion is performed on the above user behavior feature vector set to obtain a user behavior feature matrix. Among them, the above user behavior matrix can be a Boolean matrix. The row vector of the above user behavior matrix represents a user behavior feature vector. The column vector of the above user behavior matrix represents the feature vector items included in the user behavior feature vector. For example, the above behavior feature vector items can include but are not limited to at least one of the following: purchase, browse, browse duration, add to shopping cart, and favorite. The 1 in the above user behavior matrix indicates that the user behavior feature vector includes this behavior feature vector item. The 0 in the above user behavior matrix indicates that the user behavior feature vector does not include this behavior feature vector item.

[0035] In the third step, using a preset support threshold, the above user behavior feature matrix is updated to obtain an updated behavior feature matrix, and the behavior feature vector items corresponding to the column vector set included in the above updated behavior feature matrix are encoded and stored. Among them, the above preset support threshold can be a critical value of the ratio of the number of 1s in the column vector to the total number of values in the column vector that is preset in advance. The above updated behavior feature matrix can be a matrix in which the column vectors included in the updated behavior feature matrix are all column vectors with support values greater than or equal to the above preset support threshold. The above encoding and storage can be numbered and stored in ascending order according to the order of the behavior feature vector items included in the row vectors included in the user behavior feature matrix.

[0036] In the fourth step, based on the updated behavior feature matrix, the following update steps are executed:

[0037] Sub-step 1, determine the number of times the above update step has been executed.

[0038] Sub-step 2: Perform exclusive connection processing on the set of updated behavior feature column vectors included in the updated behavior feature matrix to obtain an exclusive behavior feature matrix. Among them, the exclusive behavior column vectors of the above-mentioned exclusive behavior feature matrix correspond to a user behavior feature item. The above-mentioned user behavior feature item may include multiple behavior feature vector items. For example, the exclusive behavior column vectors of the above-mentioned exclusive behavior feature matrix may be {browsing, browsing start and end time, purchase, purchase start and end time}. The above-mentioned exclusive behavior feature matrix may be a matrix including multiple user behavior feature items. The number of behavior feature vector items included in the user behavior feature item corresponding to the column vector of the exclusive behavior feature matrix may increase with the increase of the number of cycles, and each time one behavior feature vector item is added. The above-mentioned exclusive connection processing may be an exclusive connection processing that performs a logical AND operation on the updated behavior feature column vector with a smaller number and the updated behavior feature column vector with a larger number, and deletes the results with all values of 0 in the results obtained from the logical AND operation. The above-mentioned exclusive behavior feature matrix is an upper triangular matrix. Because the logical AND operation of two updated behavior feature column vectors has nothing to do with the order of operations, the updated behavior feature column vectors included in the above-mentioned exclusive behavior feature matrix are the results of the logical AND operation of the updated behavior feature column vector with a smaller number and the updated behavior feature column vector with a larger number.

[0039] As an example, the above-mentioned execution entity may first perform exclusive connection processing on any two updated behavior feature column vectors with smaller numbers and updated behavior feature column vectors with larger numbers included in the set of updated behavior feature column vectors to generate exclusive connection column vectors and obtain an exclusive connection matrix. Secondly, determine the set of exclusive connection column vectors with all values of 0 in the exclusive connection column vectors included in the above-mentioned exclusive connection matrix to obtain a target exclusive connection column vector set. Finally, delete the above-mentioned target exclusive connection column vector set from the above-mentioned exclusive connection matrix to obtain an exclusive behavior feature matrix.

[0040] Sub-step 3: Use a preset support threshold to update the above-mentioned exclusive behavior feature matrix to obtain an updated exclusive behavior feature matrix. Among them, the above-mentioned updated exclusive behavior feature matrix may be a matrix in which the ratio of the number of values of 1 included in the updated exclusive column vectors included in the updated exclusive behavior feature matrix to the total number of values included in the updated exclusive column vectors is greater than or equal to the above-mentioned preset support threshold.

[0041] Sub-step 4: In response to determining that the number of behavior feature vector items included in the updated exclusive column vectors included in the updated exclusive behavior feature matrix is less than a first preset threshold, determine the above-mentioned updated exclusive behavior feature matrix as the target behavior feature matrix. Among them, the above-mentioned first preset threshold may be the sum of the number of executed times and a second preset threshold. The above-mentioned second preset threshold may be 1.

[0042] In the fifth step, in response to determining that the number of behavior feature vector terms included in the updated mutually exclusive column vector included in the updated mutually exclusive behavior feature matrix is greater than or equal to the above-mentioned first preset threshold, the updated mutually exclusive behavior feature matrix is determined as the updated behavior feature matrix, and the sum of the number of executed times and 1 is determined as the number of executed times, so as to execute the above-mentioned update step again.

[0043] In the sixth step, according to the above-mentioned target behavior feature matrix and the preset confidence threshold, a set of target user behavior feature vectors is generated. Among them, the above-mentioned preset confidence threshold can be a threshold of an association relationship value that is preset and represents the association rule between the behavior feature vector term sets included in each target behavior column vector in the target behavior column vector set included in the target behavior feature matrix. The association rules between the behavior feature terms in the updated behavior feature terms can include: the association rules between the behavior feature terms, and the association rules between the combination of some behavior feature terms and the combination of another part of behavior feature terms in the updated behavior feature terms. The above-mentioned set of target user behavior feature vectors can be a set of feature vectors with strong association relationships. The above-mentioned strong association relationship can be that the association relationship value between each behavior feature vector term in the behavior feature vector term set included in each target behavior column vector is greater than or equal to the preset confidence threshold.

[0044] As an example, the above-mentioned execution entity can use the operation formula of the association relationship to determine the association relationship value between the behavior feature vector terms in the behavior feature vector term set included in each target behavior column vector in the target behavior column vector set included in the above-mentioned target behavior feature matrix, and obtain a set of association relationship values. Then, at least one behavior feature vector term corresponding to at least one association relationship value greater than or equal to the above-mentioned preset confidence threshold in the above-mentioned set of association relationship values is determined as the set of target user behavior feature vectors.

[0045] In the seventh step, each target user behavior feature vector in the above-mentioned set of target user behavior feature vectors is matched with a preset user abnormal behavior database to obtain a set of matching results. Among them, the above-mentioned preset user abnormal behavior database can be a feature vector database that is preset and used to store the abnormal behavior of users.

[0046] In the eighth step, at least one matching result indicating successful matching is selected from the above-mentioned set of matching results to obtain a set of target matching results.

[0047] In the ninth step, each target user behavior feature vector corresponding to the above-mentioned set of target matching results is determined as a set of abnormal user information.

[0048] The above first to ninth steps and their related content are an inventive point of the embodiments of the present disclosure, which solve the second technical problem mentioned in the background art: "When using the association rule algorithm for anomaly detection, since the association rule algorithm needs to conduct multiple comprehensive searches on the target user dataset, it consumes a large amount of computing resources, resulting in low anomaly detection efficiency, long detection time, low security of user data, and increased system load." The factors leading to low security of user data and increased system load are often as follows: When using the association rule algorithm for anomaly detection, since the association rule algorithm needs to conduct multiple comprehensive searches on the target user dataset, it consumes a large amount of computing resources, resulting in low anomaly detection efficiency and long detection time. If the above factors are solved, the security of user data can be improved and the system load can be reduced. To achieve this effect, the present disclosure first extracts features from the user behavior data included in the fused user dataset and converts the format of the user behavior feature vectors, which can improve data accuracy, reduce the data volume, and facilitate subsequent exclusive connection processing by users. Secondly, using a preset support threshold to update and encode and store the user behavior feature matrix, and deleting the column vectors included in the user behavior feature matrix that do not meet the preset support threshold can reduce the data volume and system load. Thirdly, performing cyclic exclusive connection processing and cyclic update processing on the updated behavior feature matrix can reduce the generation of implicit behavior feature vector item sets with exclusive relationships, reduce the data volume of cyclic processing, improve the efficiency of obtaining frequent item sets, and to a certain extent solve the problem of the comprehensive frequent scanning of the updated behavior feature matrix by the association rule algorithm, reducing the system load. Then, extracting the behavior feature vector item sets with strong association relationships in the target behavior feature matrix obtained by cycling can improve the accuracy of the generated target user behavior feature vector set and more accurately detect the abnormal behavior of users. Finally, performing anomaly matching on the target user behavior feature vector set and a preset user abnormal behavior database to detect whether there is abnormal behavior of users can improve the security of user data, reduce the system load, and improve the efficiency of user anomaly detection.

[0049] Step 104: Remove the abnormal user information set from the fused user dataset to obtain the post-removal user information set, which is used as the target user information set.

[0050] In some embodiments, the above execution subject may remove the above abnormal user information set from the above fused user dataset to obtain the post-removal user information set, which is used as the target user information set.

[0051] Step 105: Perform user clustering processing on the target user information set to obtain a user cluster set.

[0052] In some embodiments, the execution subject of the above network service platform adjustment method may perform user clustering processing on the above target user information set to obtain a user cluster set. Among them, the user clusters in the above user cluster set may be clusters composed of users with similar user preferences. In practice, the above execution subject may use the k-means clustering algorithm to perform user clustering processing on the above target user information set to obtain a user cluster set.

[0053] In some optional implementation manners of some embodiments, the above performing user clustering processing on the above target user information set to obtain a user cluster set may include the following steps:

[0054] First step, extract features from each target user information in the above target user information set to generate user feature vectors, and obtain a user feature vector set. Among them, the user feature vectors in the above user feature vector set may characterize the feature information of users. In practice, the above execution subject may perform the following user feature selection steps for each target user information in the above target user information set: First, use the Spearman correlation coefficient algorithm to perform attribute correlation analysis on the user attribute information set included in the above target user information to obtain an attribute correlation value set. Second, screen out at least one attribute correlation value less than or equal to a preset correlation threshold from the above attribute correlation value set to obtain a target attribute correlation value set. Among them, the above preset correlation threshold may be a preset maximum value of the attribute correlation value. For example, the above preset correlation threshold may be 0.3. Third, determine the respective user attribute information corresponding to the above target attribute correlation value set as the initial user feature attribute information set. Then, use the random forest algorithm to perform feature extraction on the above initial user feature attribute information set to obtain a user extraction feature vector set. Finally, use FA (Factor Analysis) to perform feature dimensionality reduction processing on the above user extraction feature vector set to obtain a user feature vector set.

[0055] Second step, determine the number of user clusters included in the above user cluster set as the clustering number. In practice, the above execution subject may use the elbow method to determine the number of user clusters included in the above user cluster set as the clustering number.

[0056] Third step, construct a target clustering function. Among them, the above target clustering function represents the weighted sum of the Euclidean distances and membership degree values from each user feature vector in the above user feature vector set to each initial user center feature vector in the user center feature vector set corresponding to the user cluster set. The larger the membership degree value, the closer the user feature vector corresponding to the membership degree is to the corresponding user center feature vector. The smaller the membership degree value, the farther the user feature vector corresponding to the membership degree is from the corresponding user center feature vector.

[0057] In the fourth step, according to the above-mentioned target clustering function, determine the above-mentioned number of clustering centers corresponding to the above-mentioned user feature vector set as the initial user center feature vector set. Among them, the initial user center feature vectors in the above-mentioned initial user center feature vector set can be the user feature vectors located at the central positions of the user clusters.

[0058] As an example, first, substitute the constraint condition function into the above-mentioned target clustering function to obtain a clustering constraint function. Among them, the above-mentioned constraint condition function represents that the sum of the membership degree values of each user feature vector in the above-mentioned user feature vector set to each initial user center feature vector in the above-mentioned initial user center feature vector set is 1. The above-mentioned clustering constraint function can be a function for solving the minimum value of each user feature vector in the above-mentioned user feature vector set to each initial user center feature vector in the above-mentioned initial user center feature vector set. Then, use the Lagrange multiplier method to solve the partial derivative function of the above-mentioned clustering constraint function with respect to the clustering center parameters to obtain a clustering center partial derivative function. Finally, initialize the above-mentioned clustering center partial derivative function to obtain the initial user center feature vector set.

[0059] In the fifth step, according to the above-mentioned target clustering function and the above-mentioned initial user center feature vector set, determine the membership degree values of each user feature vector in the above-mentioned user feature vector set belonging to each initial user center feature vector in the above-mentioned initial user center feature vector set as the initial membership degree values to obtain an initial membership degree value matrix.

[0060] As an example, the above-mentioned execution entity can first use the Lagrange multiplier method to solve the partial derivative function of the above-mentioned clustering constraint function with respect to the membership degree parameters to obtain a membership partial derivative function. Then, input the above-mentioned initial user center feature vector set into the above-mentioned membership partial derivative function to obtain an initial membership degree value matrix.

[0061] In the sixth step, based on the initial user center feature vector set and the initial membership degree value matrix, perform the following input steps:

[0062] The first sub-step is to update the initial user center feature vector set according to the initial membership degree value matrix to obtain an updated user center feature vector set.

[0063] As an example, the above-mentioned execution entity can perform the following cluster center update steps for each initial user center feature vector in the above-mentioned initial user center feature vector set: First, select the initial membership degree values of each user feature vector in the above-mentioned user feature vector set corresponding to the above-mentioned initial user center feature vector from the above-mentioned initial membership degree value matrix to obtain an initial membership degree value set. Second, perform a weighted sum of each user feature vector in the above-mentioned user feature vector set and the initial membership degree value corresponding to the user feature vector in the above-mentioned initial membership degree value set to obtain a weighted membership degree value. Then, determine the sum of each initial membership degree value included in the above-mentioned initial membership degree value set to obtain a cluster membership degree value. Finally, determine the ratio of the above-mentioned weighted membership degree value to the above-mentioned cluster membership degree value as the updated user center feature vector.

[0064] In the second sub-step, according to the updated user center feature vector set, determine the membership degree value of each user feature vector in the above-mentioned user feature vector set belonging to each updated user center feature vector in the updated user center feature vector set as the updated membership degree value to obtain an updated membership degree value matrix.

[0065] As an example, the above-mentioned execution entity can perform the following membership degree value update steps for each user feature vector in the above-mentioned user feature vector set and each updated user center feature vector in the above-mentioned updated user center feature vector set: First, determine the Euclidean distance value between the above-mentioned user feature vector and the above-mentioned updated user center feature vector as the user Euclidean distance value. Second, determine the Euclidean distance value between the above-mentioned user feature vector and each updated user center feature vector in the above-mentioned updated user center feature vector set as the target Euclidean distance value to obtain a target Euclidean distance value set. Then, determine the cumulative sum of the power function values with the above-mentioned user Euclidean distance value and each target Euclidean distance value in the above-mentioned target Euclidean distance value set as the base and the target value as the power exponent to obtain a user power function value. Among them, the above-mentioned target value can be the ratio of 2 to the fuzzy factor minus 1. The above-mentioned fuzzy factor can be a preset value representing the influence degree of the membership degree value on clustering. For example, the fuzzy factor can be 2. Finally, determine the ratio of 1 to the above-mentioned user power function value as the membership degree value of the user feature vector belonging to the updated user center feature vector as the updated membership degree value.

[0066] In the third sub-step, input the updated membership degree value matrix and the updated user center feature vector set into the above-mentioned target clustering function to obtain an updated clustering function value.

[0067] Fourth sub-step, in response to determining that the absolute value of the difference between the updated clustering function value and the target updated clustering function value is less than or equal to a preset difference threshold, generate a user cluster set according to the updated user center feature vector set and the updated membership degree numerical matrix. Wherein, the preset difference threshold can be a threshold of the absolute value of the difference between the updated clustering function value and the target updated clustering function value that is preset in advance. The target updated clustering function value can be the clustering function value obtained by the updated clustering function value in the previous cycle. In the first cycle, the updated clustering function value is the initial clustering function value. The initial clustering function value can be the function value obtained by inputting the initial membership degree numerical matrix and the initial user center feature vector set into the target clustering function.

[0068] As an example, the execution subject can, in response to determining that the absolute value of the difference between the updated clustering function value and the target updated clustering function value is less than or equal to the preset difference threshold. First step, for each user feature vector in the user feature vector set, perform the following determination steps: Select the membership degree value of each user feature vector belonging to each updated user center feature vector in the updated membership degree numerical matrix as the user membership degree value to obtain a user membership degree value set. Secondly, select the user membership degree value with the largest value from the user membership degree value set as the target user membership degree value. Finally, determine the updated user center feature vector corresponding to the target user membership degree value as the updated user center feature vector of the user feature vector, as the cluster center feature vector. Second step, divide the obtained cluster center feature vector set and the user feature vector set to obtain a user cluster set.

[0069] Optionally, the above method may further include the following steps:

[0070] In response to determining that the absolute value of the difference between the updated clustering function value and the target updated clustering function value is greater than the preset difference threshold, determine the updated user center feature vector set as the initial user center feature vector set, and determine the updated membership degree numerical matrix as the initial membership degree numerical matrix to execute the above input step again.

[0071] In some optional implementation manners of some embodiments, determining the above-mentioned clustering number of user clustering centers corresponding to the above-mentioned user feature vector set as the initial user center feature vector set according to the above-mentioned target clustering function may include the following steps:

[0072] First step, randomly select the above-mentioned clustering number of user feature vectors from the above-mentioned user feature vector set as the target user feature vector set.

[0073] Step 2: Generate a fitness function corresponding to the above user feature vector set. Among them, the above fitness function can represent the probability value of the target user feature vector set as the central feature vector corresponding to the above user cluster set. The above fitness function can be a function corresponding to the ratio of 1 to the above target clustering function.

[0074] Step 3: Based on the target user feature vector set, perform the following fitness screening steps:

[0075] The first sub-step: Determine the number of times the above fitness screening step has been screened and executed.

[0076] The second sub-step: Input the target user feature vector into the above fitness function to obtain an initial fitness function value set.

[0077] The third sub-step: Perform an update process on the target user feature vector set to obtain an updated user feature vector set. In practice, for each target user feature vector in the above target user feature vector set, the execution entity can determine the sum of the product of the Levy flight path and the preset step size control amount and the above target user feature vector as the updated user feature vector. Among them, the above preset step size control amount can be a preset value that conforms to the normal distribution.

[0078] The fourth sub-step: Input the updated user feature vector set into the fitness function to obtain an updated fitness function value set.

[0079] The fifth sub-step: Screen out the fitness function value with the largest function degree value from each fitness function value pair in the fitness function value pair set to obtain a target fitness function value set, where the above fitness function value pair can include: the initial fitness function value and the updated fitness function value corresponding to the initial fitness function value.

[0080] The sixth sub-step is to compare the user center probability value with the user center rejection probability value to obtain a comparison result. Among them, the above-mentioned user center probability value can be randomly generated and is a value within the range of [0, 1]. The above-mentioned user center rejection probability value can represent the probability value that the target user feature vector is not selected as the user feature vector corresponding to the user cluster center. The above-mentioned user center rejection probability value can be obtained through the following steps: First, determine the difference between the first preset rejection probability value and the second preset rejection probability value as the rejection probability difference. Among them, the above-mentioned first preset rejection probability value can be the maximum value of the user center rejection probability value set in advance. The above-mentioned second preset rejection probability value can be the minimum value of the user center rejection probability value set in advance. Second, determine the ratio of the difference between the preset screening execution threshold and the number of times of screened execution to the preset screening execution threshold as the screening execution value. Then, determine the product of the above-mentioned rejection probability difference and the above-mentioned screening execution value as the target value. Finally, determine the sum of the above-mentioned target value and the above-mentioned second preset rejection probability value as the user center rejection probability value.

[0081] The seventh sub-step, in response to determining that the comparison result indicates that the user center probability value is greater than the user center rejection probability value, randomly update each user feature vector in the target fitness function value set to obtain a target updated user feature vector set, and determine the target updated user feature vector set as the initial user center feature vector set. Among them, the above-mentioned random update can be to randomly select the number of clustering user feature vectors from the above-mentioned user feature vector set.

[0082] The eighth sub-step, in response to determining that the number of times of screened execution exceeds the preset screening execution threshold, determine the target user feature vector or the updated user feature vector corresponding to each target fitness function value in the target fitness function value set as the initial user center feature vector. Among them, the above-mentioned preset screening execution threshold can be the maximum number of executions of the above-mentioned fitness screening step set in advance.

[0083] Optionally, the above method may further include the following steps:

[0084] In response to determining that the number of times of screened execution does not exceed the above-mentioned preset screening execution threshold, determine the target user feature vector or the updated user feature vector corresponding to each target fitness function value in the target fitness function value set as the target user feature vector to obtain a target user feature vector set, and determine the sum of the number of times of screened execution and the preset value as the number of times of screened execution, and execute the above-mentioned fitness screening step again. Among them, the above-mentioned preset value can be a value set in advance. For example, the above-mentioned preset value can be 1.

[0085] Step 106: Predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set.

[0086] In some embodiments, the above-mentioned execution entity may predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set. Among them, the user churn rate value in the user churn rate value set represents the probability value of the network service platform losing users. The user churn rate value may be the average value of the user churn rates corresponding to each user included in the user cluster. The user churn rate value may be a value within the range of [0, 1]. In practice, the above-mentioned execution entity may use the XGBoost (eXtreme Gradient Boosting) model to predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set.

[0087] Step 107: Screen out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set.

[0088] In some embodiments, the above-mentioned execution entity may screen out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set.

[0089] Step 108: For each target user cluster in the target user cluster set, perform the following adjustment steps:

[0090] Step 1081: Select at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set.

[0091] In some embodiments, the above-mentioned execution entity may select at least one user comment information corresponding to the above-mentioned target user cluster from the user comment information set to obtain a target user comment information set. Among them, the above-mentioned target user cluster set may be each target user cluster corresponding to the above-mentioned target user churn rate value set.

[0092] Step 1082: Predict the user requirements for the target user comment information set to obtain a user requirement information set.

[0093] In some embodiments, the above-mentioned execution entity may perform demand prediction on the above-mentioned target user comment information set to obtain a user demand information set. Among them, the user demand information in the above-mentioned user demand information set may be some suggestions or text information for renovation proposed by the user to the network service platform. In practice, the above-mentioned execution entity may first perform sentiment analysis processing on the above-mentioned target user comment information set to obtain a sentiment tendency set. Then, filter out the comment sentences corresponding to the negative and neutral sentiment tendencies from the above-mentioned sentiment tendency set to obtain a target comment sentence set. Finally, extract the item attribute words in the above-mentioned comment sentence set to obtain an item attribute word set as the user demand information set.

[0094] In some optional implementation manners of some embodiments, the above-mentioned performing demand prediction on the above-mentioned target user comment information set to obtain a user demand information set may include the following steps:

[0095] First step, perform word segmentation processing on the text information set included in the above-mentioned target user comment information set to generate comment word segmentation groups and obtain a comment word segmentation group set. In practice, the above-mentioned execution entity may use the jieba word segmentation tool to perform word segmentation processing on the text information set included in the above-mentioned target user comment information set to generate comment word segmentation groups and obtain a comment word segmentation group set.

[0096] Second step, perform feature word recognition on the above-mentioned comment word segmentation group set to obtain a comment feature word set. Among them, the comment feature words in the above-mentioned comment feature word set may be words related to the attributes of items. For example, the above-mentioned comment feature word set may be the battery power storage of a mobile phone. In practice, the above-mentioned execution entity may use Seq2Seq4ATE (Sequence-to-Sequence for Aspect Term Extraction) to perform feature word recognition on the above-mentioned comment word segmentation group set to obtain a comment feature word set.

[0097] Third step, perform text semantic extraction on the above-mentioned comment feature word set to obtain a comment text feature vector set. Among them, the comment text feature vectors in the above-mentioned comment text feature vector set may represent the semantic information of the comment feature words and are displayed in the form of vectors. In practice, the above-mentioned execution entity may use BERT (Bidirectional Encoder Representation from Transformers) to perform text semantic extraction on the above-mentioned comment feature word set to obtain a comment text feature vector set.

[0098] Step 4: Extract features from the set of review images included in the above target user review information set to obtain a set of review image feature vectors. Among them, the review image feature vectors in the above set of review image feature vectors can represent the feature information of the review images. In practice, the above-mentioned execution entity can first use a convolutional neural network model to extract features from the review images included in the above target user review information set to obtain a set of image feature vectors. Among them, the above convolutional neural network model can be ResNet50. Then, use the DeepSentibank model to perform sentiment detection on the set of review images included in the above target user review information set to obtain a set of image sentiment text information. Then, construct a sentiment knowledge graph for the above set of image sentiment text information to obtain a sentiment knowledge graph. Finally, use a graph embedding algorithm based on random walk to embed the above sentiment knowledge graph into the above set of image feature vectors to obtain a set of review image feature vectors.

[0099] Step 5: Perform cascaded feature fusion on the above set of review text feature vectors and the above set of review image feature vectors to obtain a set of review fusion feature vectors.

[0100] Step 6: Perform weight learning on the above set of review fusion feature vectors to obtain a set of review weight feature vectors. Among them, the review weight feature vectors in the above set of review weight feature vectors can be a set of review fusion feature vectors with fusion weight values. In practice, the above-mentioned execution entity can use a self-attention mechanism to perform weight learning on the above set of review fusion feature vectors to obtain a set of review weight feature vectors.

[0101] Step 7: Input the above set of review weight feature vectors into a gated mechanism network to obtain a set of review time series feature vectors. Among them, the review time series feature vectors in the above set of review time series feature vectors can represent the time feature information of the review weight feature vectors.

[0102] Step 8: Perform prediction classification on the above set of review time series feature vectors to obtain a set of user demand information.

[0103] Step 1083: Dynamically adjust the network service platform corresponding to the user demand information set according to the user demand information set.

[0104] In some embodiments, the above-mentioned execution entity can dynamically adjust the network service platform corresponding to the above user demand information set. Among them, the above network service platform (e-commerce) can be a platform for serving users to purchase items. The above dynamic adjustment can be to adjust the user service of the network service platform, or to adjust the items included in the network service platform. The above user service can include but is not limited to at least one of the following: logistics service, value transfer recommendation service (coupon service).

[0105] As an example, the above-mentioned execution entity can determine the adjustment type that the above-mentioned network service platform needs to adjust from the above-mentioned user demand information set. Among them, the above-mentioned adjustment type can include, but is not limited to, at least one of the following: user service, item adjustment service. Then, based on the adjustment type and the user demand information set, the network service platform corresponding to the above-mentioned user demand information set is dynamically adjusted.

[0106] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: The network service platform adjustment method of some embodiments of the present disclosure can improve the efficiency of user data processing and accurately predict user needs, thereby enhancing the user experience on the network service platform. Specifically, the reasons for the relatively low user experience are as follows: Since density clustering algorithm is used for clustering and the data source is a single one, the data quality of user data in a single data source is relatively low, resulting in a low accuracy of user data analysis. In addition, the initialization of the density clustering algorithm has randomness and subjectivity, which is likely to lead to a low clustering accuracy and thus a low user experience. Based on this, the network service platform adjustment method of some embodiments of the present disclosure can, firstly, obtain user data sets and user comment information sets from different sources. Here, it can improve the integrity and accuracy of the data and is used for subsequent user clustering and user demand analysis. Secondly, perform data fusion on the above-mentioned user data sets from different sources to obtain a fused user data set. Here, data fusion can improve the consistency, integrity, and comprehensiveness of the user data set and enhance the quality of the user data set. Thirdly, perform user anomaly detection on the above-mentioned fused user data set to obtain an abnormal user information set. Here, user anomaly detection can detect abnormal user behaviors and improve the security of user data. Then, remove the above-mentioned abnormal user information set from the above-mentioned fused user data set to obtain a user information set after removal, which is used as the target user information set. Here, it can reduce the interference of useless data and the data volume, and reduce the system operation load. Subsequently, perform user clustering processing on the above-mentioned target user information set to obtain a user cluster set. Here, performing user clustering processing on the target user information set can identify the user preferences and needs of user clusters, thereby improving the user experience. After that, predict the user churn rate for each user cluster in the above-mentioned user cluster set to obtain a user churn rate value set. Here, user churn rate prediction can accurately predict the factors affecting the reduction of user experience, reduce the user churn rate, and improve the competitiveness and service level of the network service platform. Then, select at least one user churn rate value greater than or equal to a preset user churn rate threshold from the above-mentioned user churn rate value set to obtain a target user churn rate value set. Here, it is convenient for subsequent demand analysis of user clusters with a user churn rate greater than or equal to the preset user churn rate threshold to reduce the risk of user churn and improve the user experience. Finally, for each target user cluster in the target user cluster set, perform the following adjustment steps: The first step is to select at least one user comment information corresponding to the above-mentioned target user cluster from the above-mentioned user comment information set to obtain a target user comment information set, where the above-mentioned target user cluster set is each user cluster corresponding to the above-mentioned target user churn rate value set. Here, selecting from the user comment information set can improve the accuracy of user comment information, reduce the data volume for subsequent demand prediction, and reduce the system operation load.Second step, perform demand prediction on the above target user comment information set to obtain a user demand information set. Here, demand prediction can fully understand the user's needs to improve the user experience and user retention rate. Third step, according to the above user demand information set, dynamically adjust the network service platform corresponding to the above user demand information set. Here, the user experience and the service level of the network service platform can be improved. It can be seen that this network service platform adjustment method can improve the efficiency of user data processing, accurately predict user needs, improve the user experience on the network service platform, and improve the service performance of the network service platform by performing anomaly detection, clustering, churn rate identification, and demand prediction processing on user data sets from different sources.

[0107] Further reference Figure 2 , as an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a network service platform adjustment device. These device embodiments correspond to Figure 1 the method embodiments shown, and the network service platform adjustment device can be specifically applied to various electronic devices.

[0108] As Figure 2As shown in the figure, a network service platform adjustment device 200 includes: an acquisition unit 201, a data fusion unit 202, a user anomaly detection unit 203, a removal unit 204, a user clustering processing unit 205, a user churn rate prediction unit 206, a screening unit 207, and an execution unit 208. Among them, the acquisition unit 201 is configured to: acquire user data sets and user comment information sets from different sources. The data fusion unit 202 is configured to: perform data fusion on the user data sets from different sources to obtain a fused user data set. The user anomaly detection unit 203 is configured to: perform user anomaly detection on the fused user data set to obtain an abnormal user information set. The removal unit 204 is configured to: remove the abnormal user information set from the fused user data set to obtain a post-removal user information set, which is used as a target user information set. The user clustering processing unit 205 is configured to: perform user clustering processing on the target user information set to obtain a user cluster set. The user churn rate prediction unit 206 is configured to: predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set. The screening unit 207 is configured to: screen out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set. The execution unit 208 is configured to: for each target user cluster in the target user cluster set, perform the following adjustment steps: select at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set, where the target user cluster set is each user cluster corresponding to the target user churn rate value set; perform demand prediction on the target user comment information set to obtain a user demand information set; and dynamically adjust the network service platform corresponding to the user demand information set according to the user demand information set.

[0109] It can be understood that the various units described in the network service platform adjustment device 200 correspond to the respective steps in the method described in the reference Figure 1 Therefore, the operations, features, and beneficial effects described above for the method also apply to the network service platform adjustment device 200 and the units included therein, and will not be elaborated here.

[0110] Next, refer to Figure 3 , which shows a schematic structural diagram of an electronic device (for example, an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is only an example and should not impose any limitations on the functions and usage scopes of the embodiments of the present disclosure.

[0111] As Figure 3As shown, the electronic device 300 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 301, which may perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other through a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0112] Generally, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate with other devices wirelessly or wirelessly to exchange data. Although Figure 3 an electronic device 300 with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. Instead, more or fewer devices may be implemented or had. Figure 3 Each block shown in may represent one device or, as needed, multiple devices.

[0113] In particular, according to some embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, some embodiments of the present disclosure include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the methods shown in the flowcharts. In such some embodiments, the computer program may be downloaded and installed from a network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above functions defined in the methods of some embodiments of the present disclosure are executed.

[0114] It should be noted that, in some embodiments of the present disclosure, the above-mentioned computer-readable medium may be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0115] In some embodiments, the client and the server can communicate using any currently known or future-developed network protocol such as HTTP (Hyper Text Transfer Protocol), and can be interconnected with digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), the Internet (e.g., the Internet), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0116] The above computer-readable medium may be included in the above electronic device; or may exist separately without being assembled into the electronic device. The above computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device is caused to: obtain user data sets and user comment information sets from different sources; perform data fusion on the user data sets from different sources to obtain a fused user data set; perform user anomaly detection on the fused user data set to obtain an abnormal user information set; remove the abnormal user information set from the fused user data set to obtain a post-removal user information set as a target user information set; perform user clustering processing on the target user information set to obtain a user cluster set; predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set; screen out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set; for each target user cluster in the target user cluster set, perform the following adjustment steps: select at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set, where the target user cluster set is each user cluster corresponding to the target user churn rate value set; perform demand prediction on the target user comment information set to obtain a user demand information set; and dynamically adjust the network service platform corresponding to the user demand information set according to the user demand information set.

[0117] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user's computer, partially on the user's computer, execute as a stand-alone software package, execute partially on the user's computer and partially on a remote computer, or execute entirely on a remote computer or server. In the case of a remote computer, the remote computer may 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 may be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0118] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a part of code that contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0119] The units described in some embodiments of the present disclosure can be implemented in software or in hardware. The described units can also be provided in a processor. For example, it can be described as: a processor includes an acquisition unit, a data fusion unit, a user anomaly detection unit, a removal unit, a user clustering processing unit, a user churn rate prediction unit, a screening unit, and an execution unit. Among them, the names of these units do not constitute a limitation on the unit itself in some cases. For example, the acquisition unit can also be described as "the unit for acquiring user data sets and user comment information sets from different sources".

[0120] The functions described above can be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that can be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), and so on.

[0121] The above description is only some preferred embodiments of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, technical solutions formed by mutually replacing the above features with technical features having similar functions (but not limited to) disclosed in the embodiments of the present disclosure.

Claims

1. A method for adjusting a network service platform, comprising: Obtaining user data sets and user comment information sets from different sources; Performing data fusion on the user data sets from different sources to obtain a fused user data set; Performing user anomaly detection on the fused user data set to obtain an abnormal user information set; Removing the abnormal user information set from the fused user data set to obtain a post-removal user information set, which is used as a target user information set; Performing user clustering processing on the target user information set to obtain a user cluster set; Predicting the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set; Screening out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set; For each target user cluster in the target user cluster set, perform the following adjustment steps: Selecting at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set, where the target user cluster set is each user cluster corresponding to the target user churn rate value set; Performing demand prediction on the target user comment information set to obtain a user demand information set; Dynamically adjusting the network service platform corresponding to the user demand information set according to the user demand information set.

2. The method according to claim 1, wherein The performing user clustering processing on the target user information set to obtain a user cluster set includes: Extracting features for each target user information in the target user information set to generate user feature vectors, obtaining a user feature vector set; Determining the number of user clusters included in the user cluster set as the clustering number; Constructing a target clustering function; According to the target clustering function, determining the clustering number of user clustering centers corresponding to the user feature vector set as an initial user center feature vector set; According to the target clustering function and the initial user center feature vector set, determining the membership degree value of each user feature vector in the user feature vector set belonging to each initial user center feature vector in the initial user center feature vector set as an initial membership degree value, obtaining an initial membership degree value matrix; Based on the initial user center feature vector set and the initial membership degree value matrix, perform the following input steps: Updating the initial user center feature vector set according to the initial membership degree value matrix to obtain an updated user center feature vector set; According to the updated user center feature vector set, determining the membership degree value of each user feature vector in the user feature vector set belonging to each updated user center feature vector in the updated user center feature vector set as an updated membership degree value, obtaining an updated membership degree value matrix; Inputting the updated membership degree value matrix and the updated user center feature vector set into the target clustering function to obtain an updated clustering function value; In response to determining that the absolute value of the difference between the updated clustering function value and the target updated clustering function value is less than or equal to a preset difference threshold, a user cluster set is generated according to the updated user center feature vector set and the updated membership degree numerical matrix, where the target updated clustering function value is the function value obtained in the previous cycle corresponding to the updated clustering function value.

3. The method according to claim 2, wherein, The method further includes: In response to determining that the absolute value of the difference between the updated clustering function value and the target updated clustering function value is greater than the preset difference threshold, the updated user center feature vector set is determined as the initial user center feature vector set, and the updated membership degree numerical matrix is determined as the initial membership degree numerical matrix to execute the input step again.

4. The method according to claim 2, wherein The determining the clustering number of user clustering centers corresponding to the user feature vector set according to the target clustering function as the initial user center feature vector set includes: Randomly selecting the clustering number of user feature vectors from the user feature vector set as the target user feature vector set; Generating a fitness function corresponding to the user feature vector set; Based on the target user feature vector set, performing the following fitness screening steps: Determining the number of times the fitness screening step has been screened and executed; Inputting the target user feature vector into the fitness function to obtain an initial fitness function value set; Performing an update process on the target user feature vector set to obtain an updated user feature vector set; Inputting the updated user feature vector set into the fitness function to obtain an updated fitness function value set; Screening out the fitness function value with the largest function degree value from each fitness function value pair in the fitness function value pair set to obtain a target fitness function value set, where the fitness function value pair includes: an initial fitness function value and an updated fitness function value corresponding to the initial fitness function value; Comparing the user center probability value with the user center discard probability value to obtain a comparison result; In response to determining that the comparison result indicates that the user center probability value is greater than the user center discard probability value, randomly updating each user feature vector corresponding to the target fitness function value set to obtain a target updated user feature vector set, and determining the target updated user feature vector set as the initial user center feature vector set; In response to determining that the number of times the screening has been executed exceeds a preset screening execution threshold, determining each target user feature vector or updated user feature vector corresponding to each target fitness function value in the target fitness function value set as the initial user center feature vector.

5. The method according to claim 4, wherein, The method further includes: In response to determining that the number of times the screening has been executed does not exceed the preset screening execution threshold, determining each target user feature vector or updated user feature vector corresponding to each target fitness function value in the target fitness function value set as the target user feature vector to obtain a target user feature vector set, determining the sum of the number of times the screening has been executed and a preset value as the number of times the screening has been executed, and executing the fitness screening step again.

6. The method according to claim 1, wherein, The predicting user demand information set by performing demand prediction on the target user comment information set includes: Perform word segmentation on the text information set included in the target user comment information set to generate comment word segmentation groups, and obtain a comment word segmentation group set; Perform feature word recognition on the comment word segmentation group set to obtain a comment feature word set; Perform text semantic extraction on the comment feature word set to obtain a comment text feature vector set; Perform feature extraction on the comment image set included in the target user comment information set to obtain a comment image feature vector set; Perform cascaded feature fusion on the comment text feature vector set and the comment image feature vector set to obtain a comment fusion feature vector set; Perform weight learning on the comment fusion feature vector set to obtain a comment weight feature vector set; Input the comment weight feature vector set into a gated mechanism network to obtain a comment time series feature vector set; Perform prediction classification on the comment time series feature vector set to obtain a user demand information set.

7. A network service platform adjustment device, comprising: An acquisition unit configured to acquire user data sets and user comment information sets from different sources; A data fusion unit configured to perform data fusion on the user data sets from different sources to obtain a fused user data set; A user anomaly detection unit configured to perform user anomaly detection on the fused user data set to obtain an abnormal user information set; A removal unit configured to remove the abnormal user information set from the fused user data set to obtain a post-removal user information set as a target user information set; A user clustering processing unit configured to perform user clustering processing on the target user information set to obtain a user cluster set; A user churn rate prediction unit configured to predict the user churn rate for each user cluster in the user cluster set to obtain a user churn rate value set; A screening unit configured to screen out at least one user churn rate value greater than or equal to a preset user churn rate threshold from the user churn rate value set to obtain a target user churn rate value set; An execution unit configured to, for each target user cluster in the target user cluster set, perform the following adjustment steps: select at least one user comment information corresponding to the target user cluster from the user comment information set to obtain a target user comment information set, where the target user cluster set is each target user cluster corresponding to the target user churn rate value set; Perform demand prediction on the target user comment information set to obtain a user demand information set; and dynamically adjust the network service platform corresponding to the user demand information set according to the user demand information set.

8. An electronic device, comprising: One or more processors; A storage device having stored thereon one or more programs, When the one or more programs are executed by the one or more processors, causing the one or more processors to implement the method according to any one of claims 1-6.

9. A computer-readable medium having a computer program stored thereon, wherein, The computer program, when executed by a processor, implements the method according to any one of claims 1-6.

Images