User credibility calculation method and device based on time sequence, medium and equipment

Abstract

The invention relates to a user credibility calculation method and device based on a time sequence, a medium and equipment, and relates to the technical field of data processing, and the method comprises the steps: carrying out the clustering analysis of each target social platform according to platform behavior feature data, and obtaining a first weight corresponding to each target social platform; performing clustering analysis on each target user according to the individual behavior feature data to obtain a second weight corresponding to each target user, and determining a behavior weight of a single behavior of the corresponding target user according to the second weight; determining an initial credibility score of the corresponding target user according to the behavior data of the single target user on each target social platform, each first weight and each corresponding behavior weight; and determining a final credibility score of the corresponding target user according to the initial credibility score and the future credibility score corresponding to the same target user. The method and the device have the effect of improving the accuracy of credibility evaluation.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, medium and equipment for calculating user reputation based on time series. Background Art

[0002] With the rapid development of mobile Internet, social platforms play an increasingly important role in modern life. The behaviors of users on different social platforms, such as sending messages, liking, commenting, sending voice, sending pictures, sending videos, sending red envelopes and transferring money, can reflect their participation and activity on social platforms. In order to better maintain the healthy development of social platforms and make users have a better experience on social platforms, social platforms usually conduct credibility assessments on users to measure the credibility and behavior of users on social platforms. On this basis, users with high credibility are more likely to gain the trust of other users, thereby promoting healthier and more positive social interactions; in addition, through the assessed credibility, the platform can identify users who behave improperly or publish false information, thereby better maintaining the order of the platform and user experience.

[0003] At present, the usual way to evaluate the reputation of users is to use a simple scoring mechanism to score users based on the frequency of specific behaviors on a single social platform, such as scoring users based on the number of likes or comments they make on a social platform. The more likes, the higher the reputation score. However, the diversity of users' behaviors that reflect their reputation on social platforms is high, and there are large differences in behavior on different social platforms. In this way, evaluating the reputation of users based on the frequency of a specific behavior on a single social platform cannot fully reflect the comprehensive behavior of users on multiple social platforms, resulting in poor accuracy of reputation evaluation. Summary of the invention

[0004] In order to improve the accuracy of reputation evaluation, the present application provides a user reputation calculation method, device, medium and equipment based on time series.

[0005] In a first aspect of the present application, a method for calculating user reputation based on time series is provided, which specifically includes: Obtaining behavior data of each target user in at least one target social platform; Determining platform behavior feature data of a single target social platform according to the behavior data, and performing cluster analysis on each target social platform according to the platform behavior feature data to obtain a first weight corresponding to each target social platform; Determine individual behavior feature data of a single target user on each target social platform based on the behavior data, perform cluster analysis on each target user based on the individual behavior feature data to obtain a second weight corresponding to each target user, and determine the behavior weight of a single behavior of the corresponding target user based on the second weight; Determining an initial credibility score of a corresponding target user according to the behavior data of a single target user on each of the target social platforms, each of the first weights and the corresponding behavior weights; Based on a preset time series model, at least one future credibility score of a single target user in the target social platform is obtained, and according to the initial credibility score and the future credibility score corresponding to the same target user, the final credibility score of the corresponding target user is determined. The time series model is used to predict the credibility score of the target user after the current time based on the credibility score of the target user before the current time, and the future credibility score is an instant credibility score within a preset time after the current time.

[0006] By adopting the above technical solution, after obtaining the behavior data of the target users in each target social platform, the specific situation of the user behavior characteristics on each target social platform, that is, the platform behavior characteristic data, is analyzed and determined based on the behavior data, and cluster analysis is performed on each target social platform accordingly, and then based on the clustering results, the first weight of each target social platform matching is determined specifically, thereby reflecting the user behavior characteristics on the target social platform. Similarly, the second weight of each target user and the behavior weight of a single behavior can be determined to reflect the behavior characteristics of different groups on the social platform. Further, for a single target user, the behavior data on each target social platform, the first weight of each target social platform and each behavior weight are summarized and considered to determine its initial credibility score, which preliminarily reflects its participation or credibility on the target social platform. Finally, each initial credibility score is adjusted and optimized to obtain the final credibility score, thereby improving the accuracy of the credibility assessment.

[0007] Optionally, performing cluster analysis on each of the target social platforms according to the platform behavior characteristic data to obtain a first weight corresponding to each of the target social platforms specifically includes: Based on the platform behavior feature data corresponding to a single target social platform, construct a platform feature vector of the corresponding target social platform; Performing cluster analysis on each of the platform feature vectors using a preset clustering algorithm to obtain a corresponding first clustering result; According to the first clustering result, all the target social platforms are divided into at least one first clustering group, and according to the first clustering group to which each target social platform belongs, a basic weight of the corresponding target social platform is determined; The basic weight corresponding to a single target social platform is adjusted and optimized to obtain an adjusted weight, and each adjusted weight is normalized to obtain a first weight corresponding to each target social platform.

[0008] By adopting the above technical solution, cluster analysis is performed on the feature vectors of each platform to obtain the first clustering result, so as to classify the feature vectors of each platform according to the similarity of the behavioral characteristics of the platform users. Then, the behavioral characteristics of each group are analyzed to determine the behavioral patterns of the target users on the target social platforms corresponding to the feature vectors of each platform in each group, and then the target social platforms are divided into at least one first cluster group according to the behavioral model. Furthermore, according to the first cluster group to which they belong, the basic weight of each target social platform can be determined more reasonably, and then the final first weight can be determined, so as to facilitate the subsequent accurate calculation of the credibility score of the target user.

[0009] Optionally, the adjusting and optimizing the basic weight corresponding to the single target social platform to obtain the adjusted weight specifically includes: Determine, according to the platform behavior characteristic data of a single target social platform, a weight adjustment factor of at least one dimension of the corresponding target social platform, wherein the weight adjustment factor reflects the degree of influence of the platform behavior characteristic data of different dimensions on the basic weight; A weighted sum is performed on each of the weight adjustment factors to obtain a sum result, and the sum result and the basic weight of the corresponding target social platform are summed to obtain an adjusted weight.

[0010] By adopting the above technical solution, after the basic weight of a single target social platform is determined, since the basic weight is determined based on the behavioral commonalities of the second clustering group to which it belongs, it is necessary to adjust and optimize the basic weight corresponding to the target social platform according to the user behavior characteristics presented by the target social platform itself, obtain the adjusted weight, and normalize each adjusted weight to obtain the first weight corresponding to each target social platform, thereby making the platform weight of each target social platform more accurate and reasonable.

[0011] Optionally, performing cluster analysis on each of the target users according to the individual behavior characteristic data to obtain a second weight corresponding to each target user specifically includes: Based on the individual behavior feature data of a single target user, construct an individual feature vector of the corresponding target user; Performing cluster analysis on each of the individual feature vectors using a preset clustering algorithm to obtain a corresponding second clustering result; According to the second clustering result, all the target users are divided into at least one second clustering group, and according to the second clustering group to which each target user belongs, a second weight of the corresponding target user is determined.

[0012] By adopting the above technical solution, cluster analysis is performed on each individual feature vector to obtain the second clustering result, so as to classify each individual feature vector according to feature similarity, and analyze the user behavior characteristics of each cluster, so as to determine the behavior pattern of the target user corresponding to the individual feature vector in each cluster, and divide all target users into at least one second cluster group. Finally, according to the behavior characteristics of the second cluster group to which the target user belongs, the second weight corresponding to the target user is determined, so as to facilitate the subsequent more accurate determination of the reputation score.

[0013] Optionally, determining the behavior weight of a single behavior of the corresponding target user according to the second weight specifically includes: Determine, according to the behavior data of a single target user in each target social platform, at least one evaluation index weight for each behavior of the corresponding target user, wherein the evaluation index weight is an index weight for evaluating the importance of a single behavior of the target user; The weights of the evaluation indicators are summed to obtain the sum of the weights of the corresponding behaviors; The sum of the weights is multiplied by the second weight of the corresponding target user to obtain the behavior weight of each behavior of the corresponding target user.

[0014] By adopting the above technical solution, after determining the second weight of the target user according to the second clustering group to which he belongs, in order to more accurately determine the behavioral weight of a single behavior of the target user and make the subsequent credibility evaluation more accurate, it is necessary to consider the personalized behavioral performance of the target user himself and evaluate the importance of each of his behaviors for the credibility evaluation. Then, the sum of the weights of each evaluation indicator weight of each behavior of the target user is determined to characterize the importance of the behavior. Finally, the sum of the weights is multiplied by the corresponding second weight of the target user, so as to better integrate the common behavioral performance of the group to which the target user belongs and the personalized behavioral performance of the target user, and more accurately determine the behavioral weight of a single behavior.

[0015] Optionally, determining the initial credibility score of the corresponding target user according to the behavior data of the single target user on each of the target social platforms, each of the first weights and the corresponding behavior weights specifically includes: Determine the frequency of a single behavior of a corresponding target user on a single target social platform according to the behavior data of the single target user on each target social platform; According to each of the first weights, the behavior frequency and behavior weight corresponding to a single behavior of a single target user in a single target social platform, the initial reputation score of the corresponding target user is determined by a preset reputation score calculation formula, and the reputation score calculation formula is: ; In the formula, S represents the initial reputation score, j represents the jth target social platform, M represents the total number of target social platforms, i represents the i-th behavior, and N represents the total number of different types of behaviors of the target user on the target social platform. represents the first weight of the jth target social platform, represents the behavior weight corresponding to the target user's ith behavior, represents the behavior frequency of the target user's i-th behavior in the j-th target social platform, represents the preset error damping factor, represents the preset variance damping factor, represents the error coefficient of the target user's i-th behavior in the j-th target social platform, Represents the variance coefficient of the target user's i-th behavior in the j-th target social platform.

[0016] Optionally, determining the final reputation score of the corresponding target user according to the initial reputation score and the future reputation score corresponding to the same target user specifically includes: Determine a corresponding reputation score error according to the future reputation score of a single target user and the corresponding actual reputation score, wherein the actual reputation score is an actual value of the reputation score within a preset time after the current time; Normalizing each of the reputation score errors to obtain processed score errors, and calculating a target mean and a target standard deviation of all processed score errors; Substitute the target mean, target standard deviation and initial reputation score corresponding to a single target user into a preset reputation score adjustment formula to obtain the final reputation score of the corresponding target user. The reputation score adjustment formula is: ; In the formula, Z represents the final reputation score of the target user, S represents the initial reputation score of the target user, represents the target mean corresponding to the target user, represents the target standard deviation corresponding to the target user, represents the weight factor of the target mean, A weighting factor representing the target standard deviation.

[0017] By adopting the above technical solution, the target mean and target standard deviation of all processed scoring errors are determined, so as to better characterize the size and fluctuation of the error in the reputation score of the target user. Furthermore, the target mean and target standard deviation of a single target user are substituted into the reputation score adjustment formula, and the initial reputation score is reasonably adjusted based on the error that may occur when evaluating the reputation of a single target user, so as to ensure the reliability and accuracy of the reputation score of the target user.

[0018] In a second aspect of the present application, a user reputation calculation device based on time series is provided, which specifically includes: A data acquisition module, used to acquire behavior data of each target user in at least one target social platform; A first weight module, configured to determine platform behavior characteristic data of a single target social platform according to the behavior data, and to perform cluster analysis on each of the target social platforms according to the platform behavior characteristic data to obtain a first weight corresponding to each of the target social platforms; A second weight module is used to determine individual behavior feature data of a single target user on each target social platform based on the behavior data, perform cluster analysis on each target user based on the individual behavior feature data, obtain a second weight corresponding to each target user, and determine a behavior weight of a single behavior of the corresponding target user based on the second weight; A reputation scoring module, used to determine an initial reputation score of a corresponding target user based on the behavior data of a single target user on each of the target social platforms, each of the first weights and the corresponding behavior weights; The score adjustment module is used to obtain at least one future reputation score of a single target user in the target social platform based on a preset time series model, and determine the final reputation score of the corresponding target user according to the initial reputation score and the future reputation score corresponding to the same target user. The time series model is used to predict the reputation score of the target user after the current time based on the reputation score of the target user before the current time, and the future reputation score is an instant reputation score within a preset time after the current time.

[0019] By adopting the above technical solution, the data acquisition module obtains the behavior data of the target user, and then the first weight module determines the platform behavior feature data of the single target social platform based on the behavior data, and performs cluster analysis on each target social platform to obtain the first weight of each target social platform; the second weight module determines the individual behavior feature data of the single target user, and determines the second weight of each target user and the behavior weight of a single behavior, and then the reputation scoring module determines the initial reputation score of the target user, and finally the score adjustment module adjusts the initial reputation score to obtain the final reputation score.

[0020] In a third aspect of the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is loaded and executed by a processor, the method steps described in any one of the first aspects are performed.

[0021] In a fourth aspect of the present application, an electronic device is provided, specifically comprising: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the processor is used to load and execute the computer program stored in the memory so that the electronic device performs the method as described in any one of the first aspects.

[0022] In summary, the present application includes at least one of the following beneficial technical effects: based on behavioral data, the specific situation of user behavior characteristics on each target social platform is analyzed and determined, that is, the platform behavior characteristic data, and each target social platform is clustered accordingly, and then based on the clustering results, the first weight of each target social platform matching is determined specifically, thereby reflecting the user behavior characteristics on the target social platform. Similarly, the second weight of each target user and the behavioral weight of a single behavior can be determined to reflect the behavioral characteristics of different groups on social platforms. Furthermore, for a single target user, the behavioral data on each target social platform, the first weight of each target social platform, and each behavioral weight are summarized and considered to determine its initial credibility score, which preliminarily reflects its participation or credibility on the target social platform. Finally, each initial credibility score is adjusted and optimized to obtain a final credibility score, thereby improving the accuracy of the credibility assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of a method for calculating user reputation based on time series provided in an embodiment of the present application; Figure 2 It is a structural diagram of a user reputation calculation device based on time series provided in an embodiment of the present application.

[0024] Explanation of the accompanying drawings: 11. Data acquisition module; 12. First weight module; 13. Second weight module; 14. Reputation scoring module; 15. Scoring adjustment module. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0026] In the description of the embodiments of the present application, words such as "illustrative", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "illustrative", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "illustrative", "for example" or "for example" is intended to present related concepts in a concrete way.

[0027] In the description of the embodiments of the present application, the term "and / or" is only a kind of association relationship describing the associated objects, indicating that there may be three kinds of relationships, for example, A and / or B, which can represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0028] See also Figure 1 The present application embodiment discloses a flow chart of a method for calculating user reputation based on time series, which can be implemented by a computer program or run on a user reputation calculation device based on a time series based on a von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S101: Obtaining behavior data of each target user in at least one target social platform.

[0029] Specifically, in an embodiment of the present application, the target social platform may be an application of social attributes, and in other embodiments, the target social platform may also be a webpage platform of social attributes. The target user is a user who needs to evaluate the credibility in the social platform and determine the participation and activity on the social platform. The behavior data reflects the specific situation information of multiple behaviors of the target user in each target social platform. The behavior of the target user in the target social platform includes but is not limited to sending text messages, sending voice messages, sending pictures, sending videos, sending red envelopes, transferring money, publishing articles, and liking comments. The behavior data includes but is not limited to the number of times the target user sends text messages, the number of times voice messages are sent, the number of times pictures are sent, the number of times videos are sent, the number and amount of red envelopes, the number and amount of transfers, the number of times articles are published, and the number of times comments are liked. It should be noted that the behavior data can cover different time periods, and the frequency of the behavior of the target user is recorded in fine granularity such as hourly, daily, and weekly. In other embodiments, the behavior data can also cover a single time period.

[0030] In addition, the execution subject of a user reputation calculation method based on time series disclosed in an embodiment of the present application is a server, and the server can be an independent physical server or a server cluster composed of multiple physical servers. Furthermore, a feasible way to obtain behavioral data is to obtain the behavioral data of target users in each target social platform through the API interface or log file provided by each target social platform. In other embodiments, after obtaining the data information of each behavior of the target user through the API interface or log file provided by each target social platform, all the data information is preprocessed to obtain the final processed behavioral data. Preprocessing includes data cleaning processing and data standardization processing. Among them, data cleaning processing aims to remove or correct various problems in the original data to ensure data quality. Data standardization processing aims to eliminate the scale differences of different behavioral data, and is specifically processed through Z-score standardization. This is a prior art and will not be repeated here.

[0031] S102: Determine platform behavior characteristic data of a single target social platform based on the behavior data, and perform cluster analysis on each target social platform based on the platform behavior characteristic data to obtain a first weight corresponding to each target social platform.

[0032] Specifically, after the behavior data of each target user is determined, it is necessary to determine the platform behavior characteristic data of the corresponding target social platform according to the behavior data of each target user in a single target social platform. The platform behavior characteristic data is used to characterize the behavior characteristics of each target user in the target social platform. In an embodiment of the present application, a feasible determination method is: counting the average number of target users who are active every day on a single target social platform to obtain the corresponding daily average number of active users. Further, counting the average number of target users who are active every month on a single target social platform to obtain the corresponding monthly active users. Count the average number of operations of each target user on a single target social platform, that is, the ratio of the total number of behaviors of all target users in a single target social platform (the sum of the number of messages sent, the number of likes and comments, etc.) to the total number of target users, thereby reflecting the activity level of a single target user on the target social platform. Finally, these statistical data are determined as the platform behavior characteristic data of the target social platform. In other embodiments, the average amount of fund transactions (such as sending red envelopes or transferring money) conducted on a single target social platform every day can also be counted to obtain the average daily transaction amount, and it is also determined as the platform behavior characteristic data of the target social platform.

[0033] Further, for a single target social platform, the corresponding platform behavior feature data is constructed as a platform feature vector through principal component analysis. In other embodiments, the platform feature vector can also be constructed by singular value decomposition. For example, the platform feature vector corresponding to the target social platform A is expressed as: (average number of daily active users, monthly active users, average number of operations...). Then, the platform feature vectors corresponding to each target social platform are clustered and analyzed by a preset clustering algorithm to obtain the corresponding first clustering result, wherein the clustering algorithm can use K-Means clustering. In other embodiments, the clustering algorithm can also use hierarchical clustering. The first clustering result includes several groups or clusters after clustering the platform feature vectors of each target social platform and the platform feature vectors contained in each group. Further, the characteristics of each group are analyzed by a descriptive statistical analysis method, and the behavior pattern of the target user on the target social platform corresponding to the platform feature vector in each group is determined, and finally all target social platforms are divided into at least one first clustering group. For example, the first clustering result contains two clusters. The common characteristics of the platform feature vectors in cluster 1 are: high daily active users, high interaction frequency, and low transaction amount. The common characteristics of the platform feature vectors in cluster 2 are: low daily active users, low interaction frequency, and high transaction amount. Then the final clustering groups are: a group with high daily active users, high interaction frequency, and low transaction amount, and a group with low daily active users, low interaction frequency, and high transaction amount.

[0034] Furthermore, by using indicators such as information gain in the decision tree, the importance of each first clustering group in terms of user activity, interaction frequency, etc. is evaluated to determine the basic weight of each first clustering group, and then the basic weight of the corresponding target social platform is determined according to the first clustering group to which each target social platform belongs. After determining the basic weight of a single target social platform based on the basic weight of the first clustering group, it is necessary to adjust and optimize the basic weight corresponding to the target social platform according to the user behavior characteristics presented by the target social platform itself, obtain the adjusted weight, and normalize each adjusted weight to obtain the first weight corresponding to each target social platform, so that the platform weight of each target social platform is more accurate and reasonable. The larger the basic weight, the more important the characteristics of the corresponding first clustering group are to the activity of the target social platform.

[0035] Among them, the basic weight corresponding to the target social platform is adjusted and optimized. One achievable implementation method is: according to the platform behavior characteristic data of a single target social platform, the weight adjustment factor of at least one dimension of the target social platform is determined. The weight adjustment factor reflects the influence of the platform behavior characteristic data of different dimensions on the basic weight. The larger the weight adjustment factor, the greater the influence on the basic weight. Exemplarily, the average daily number of active users is obtained from the platform behavior characteristic data of a single target social platform, and the ratio of this target number of active users to the sum of the average daily number of active users of all target social platforms is determined as the weight adjustment factor of the user activity dimension. The larger the ratio, the higher the user activity of the target social platform, and the corresponding basic weight is correspondingly higher. In other embodiments, the weight adjustment factor of a single target social platform may also include a weight adjustment factor of the platform influence dimension, which is specifically determined as follows: the number of types of people involved in the target users of a single target social platform is counted. The more the number, the easier it is for the target social platform to attract more different people to use it, the greater the platform influence, the larger the weight adjustment factor of the platform influence dimension, and the corresponding basic weight is correspondingly higher.

[0036] Furthermore, each weight adjustment factor is weighted and summed to obtain a sum result, and finally the sum result is summed with the basic weight of the corresponding target social platform to obtain the adjusted weight corresponding to the target social platform. Among them, the weight of each weight adjustment factor when weighted and summed can be determined according to indicators such as information gain in the decision tree. This is a prior art and will not be repeated here. In other embodiments, the basic weight of a single target social platform can also be directly determined as the final first weight.

[0037] S103: Determine the individual behavioral characteristic data of a single target user on each target social platform based on the behavioral data, perform cluster analysis on each target user based on the individual behavioral characteristic data, obtain a second weight corresponding to each target user, and determine the behavioral weight of a single behavior of the corresponding target user based on the second weight.

[0038] Specifically, after the first weight of each target social platform is determined, the behavior data of a single target user is screened from all the behavior data, and then the individual behavior feature data of the single target user on each target social platform is determined. The individual behavior feature data is used to characterize the behavioral characteristics of the single target user on each target social platform. A feasible way to determine the individual behavior feature data is to sum the number of single behaviors of the target user on each target social platform to obtain the behavior frequency of the single behavior. Furthermore, based on the behavior frequency of various behaviors, the corresponding mean, variance, median, etc. are calculated. These data are determined as the individual behavior feature data of the target user. In other embodiments, the Pearson correlation coefficient between different behaviors of the target user can also be used to determine the correlation between different behaviors, and this correlation can also be used as individual behavior feature data. Then, based on the individual behavior feature data of the target user, the individual feature vector to which it belongs is constructed through the principal component analysis method.

[0039] Furthermore, by using a preset clustering algorithm, cluster analysis is performed on the individual feature vectors of each target user to obtain a corresponding second clustering result, wherein the second clustering result includes several groups or clusters after clustering the individual feature vectors of each target user and the individual feature vectors contained in each group. Furthermore, by using a descriptive statistical analysis method, the characteristics of each group are analyzed according to the second clustering result, and the behavior pattern of the target user corresponding to the individual feature vector in each group is determined, and finally all target users are divided into at least one second clustering group. Among them, each second clustering group includes at least one target user, and the behavior characteristics of each target user in each second clustering group are relatively similar. Exemplarily, the behavior pattern of group 3 is: high message sending frequency, high like frequency, and low video sending frequency. The behavior pattern of group 4 is: low message sending frequency, low like frequency, and high video sending frequency. Then the target users corresponding to the individual feature vectors in cluster 3 are divided into a second cluster group and named as the high interactive user group; the target users corresponding to the individual feature vectors in cluster 4 are divided into another second cluster group and named as the high content output user group. Finally, through the information gain and other indicators in the decision tree, the importance of each second cluster group to user activity, interaction frequency, etc. is analyzed, and then the basic weight of each second cluster group is determined, and the basic weight of the second cluster group to which the single target user belongs is determined as the second weight of the corresponding target user. The larger the second weight, the more prominent the corresponding target user is in terms of activity and participation on the target social platform.

[0040] After the second weight of a single target user is determined, this second weight is directly determined as the behavior weight of each behavior of the target user. In other embodiments, the number of single behaviors can also be screened out from the behavior data of the target user in each target social platform and summed to obtain the total number of times. The ratio of this total number to the total number of times of all behaviors is determined as the evaluation index weight of the behavior, wherein the evaluation index weight is the index weight for evaluating the importance of a single behavior of the target user. In other embodiments, the ratio of the number of time periods in which a single behavior of the target user occurs to the total number of time periods in which all behaviors occur can also be counted to determine the evaluation index weight of the behavior, thereby measuring the importance of a single behavior of a single target user. Then sum the weights of each evaluation index to obtain the sum of the weights of the corresponding behavior. Finally, multiply this sum of weights by the second weight of the corresponding target user to obtain the behavior weight of the corresponding single behavior of the target user.

[0041] S104: Determine an initial credibility score of a corresponding target user according to the behavior data of a single target user on each target social platform, each first weight and each corresponding behavior weight.

[0042] Specifically, from the behavior data of a single target user on each target social platform, the behavior frequency of a single behavior of the target user on the single target social platform is screened out, and then the first weight of each target social platform, the behavior frequency and behavior weight corresponding to a single behavior of the single target user on the single target social platform are substituted into the preset reputation score calculation formula to obtain the initial reputation score of the target user. The reputation score calculation formula is: ; In the formula, S represents the initial reputation score, j represents the jth target social platform, M represents the total number of target social platforms, i represents the i-th behavior, and N represents the total number of different types of behaviors of the target user on the target social platform. represents the first weight of the jth target social platform, represents the behavior weight corresponding to the target user's ith behavior, represents the behavior frequency of the target user's i-th behavior in the j-th target social platform, represents the preset error damping factor, represents the preset variance damping factor, represents the error coefficient of the target user's i-th behavior in the j-th target social platform, Represents the variance coefficient of the target user's i-th behavior in the j-th target social platform. It should be noted that the error damping factor is used to adjust the impact of the error coefficient on the score and smooth out the fluctuations in the score caused by abnormal behavior. In the embodiment of the present application, the error damping factor can be set to 0.2, indicating that the error coefficient has a 20% impact on the score. The variance damping factor is used to adjust the impact of the variance coefficient on the score, reflecting the stability and volatility of the target user's behavior. The variance damping factor can be set to 0.1, indicating that the variance coefficient has a 10% impact on the score. In addition, Reflects the volatility or stability of the i-th behavior in the j-th target social platform. is used to control the sensitivity of the variance to the score, both of which are set based on the historical behavior data of the target users and human experience.

[0043] S105: Based on a preset time series model, at least one future reputation score of a single target user in the target social platform is obtained, and according to the initial reputation score and the future reputation score corresponding to the same target user, a final reputation score of the corresponding target user is determined.

[0044] Specifically, after the initial reputation score of a single target user is determined, in order to improve the accuracy of the target user's reputation determination, it is necessary to verify and optimize the initial reputation score. A feasible implementation method is: based on the behavior data of a single target user within a preset time before the current time, at least one historical reputation score is determined. The specific determination method can be referred to step S104, which will not be repeated here. Then, each historical reputation score is sorted in chronological order to obtain time series data, and then this time series data is input into a preset time series model, which is used to predict the reputation score after the current time based on the reputation score of the target user before the current time, and finally obtain at least one future reputation score of the single target user in the target social platform, that is, the instant reputation score within the preset time after the current time. Among them, the time series model is a trained ARIMA model, and in other embodiments, it can also be a trained LSTM neural network model. The training process is briefly described as follows: the reputation scores before and after the preset time nodes of each user sample are sorted into time series data, respectively, and input into the model for training, and the reverse gradient algorithm is used to optimize the model parameters, and finally the time series model is obtained. This is prior art and will not be described in detail here.

[0045] Furthermore, the future reputation score of a single target user and the corresponding actual reputation score are processed by difference, and the absolute value of the difference is taken to finally obtain the reputation score error. The larger the reputation score error is, the more unstable the reputation score of the target user in the target social platform is. Among them, the actual reputation score is the actual value of the reputation score at the same time node as the future reputation score within a preset time after the current time. Further, each reputation score error is normalized to obtain a processed score error, and then the target mean and target standard deviation of all processed score errors are calculated, and finally the target mean, target standard deviation and initial reputation score corresponding to the single target user are substituted into the preset reputation score adjustment formula to obtain the final reputation score of the target user. Among them, the reputation score adjustment formula is: ; In the formula, Z represents the final reputation score of the target user, S represents the initial reputation score of the target user, represents the target mean corresponding to the target user, represents the target standard deviation corresponding to the target user, represents the weight factor of the target mean, Represents the weight factor of the target standard deviation. It should be noted that in this process, the time series model is used to predict the reputation score, so as to determine the error of the reputation score, and then reasonably adjust the initial reputation score of a single target user to ensure the reliability and accuracy of the reputation score of the target user.

[0046] In other embodiments, the historical operation types of the target user on a single target social platform are obtained, the first occurrence number of each historical operation type is counted, and a first number of historical operation types are selected from each historical operation type in descending order of the first occurrence number to determine as the target operation type, that is, the operation type that is easy to appear on the target social platform; the historical occurrence period of a single target operation type of the target user is obtained, the second occurrence number of each historical operation type is counted, and a second number of historical occurrence periods are selected from each historical occurrence period in descending order of the second occurrence number to determine as the target occurrence period of the corresponding target operation type, that is, the period that is easy to appear. Further, for a single target social platform, a third weight of the target operation type and a fourth weight of each corresponding target occurrence period are determined, the third weight being the ratio of the first occurrence number of each target operation type to the sum of the first occurrence numbers of all target operation types, and the fourth weight being the ratio of the second occurrence number of the single target occurrence period corresponding to the target operation type to the sum of the second occurrence numbers of all corresponding target occurrence periods.

[0047] Furthermore, if the final credibility score of the target user is lower than a preset score threshold, it means that the target user's behavior on the social platform is likely to have a negative impact and affect the healthy operation of the social platform. Then, for a single target social platform, the target operation type that exists at the current time in the corresponding target appearance time period is determined as the key operation type, and the target appearance time period that exists at the current time is determined as the key appearance time period. The product of the third weight of each key operation type and the fourth weight of the corresponding key appearance time period is calculated. The larger the product is, the more likely the target user will have the corresponding key operation type on the target social platform at the current time and in a short time thereafter. The products are summed to obtain the sum of the products of the corresponding target social platforms. The larger the sum of the products is, the more likely the target user will have an operation on the corresponding target social platform at the current time and in a short time thereafter. The maximum sum of products is selected from the sums of all products, and the target social platform corresponding to the maximum sum of products (the social platform where the target user is most likely to perform operations) is determined as the key social platform. The key operation type corresponding to the maximum product among the products corresponding to the key social platform is determined as the monitored operation type. Furthermore, starting from the current time, real-time key monitoring of the monitoring operation type of this target user is carried out on the key social platforms to prevent the target user's operations on the key social platforms from having a negative impact on the platforms.

[0048] The implementation principle of the user reputation calculation method based on time series in the embodiment of the present application is: based on the behavior data, analyze and determine the specific situation of the user behavior characteristics on each target social platform, that is, the platform behavior characteristic data, and perform cluster analysis on each target social platform accordingly, and then based on the clustering results, determine the first weight of each target social platform matching in a targeted manner, thereby reflecting the user behavior characteristics on the target social platform. Similarly, the second weight of each target user and the behavior weight of a single behavior can be determined to reflect the behavioral characteristics of different groups on the social platform. Further, for a single target user, summarize and consider its behavior data on each target social platform, the first weight of each target social platform, and each behavior weight, and determine its initial reputation score, which preliminarily reflects its participation or credibility on the target social platform. Finally, each initial reputation score is adjusted and optimized to obtain a final reputation score, thereby improving the accuracy of the reputation assessment.

[0049] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0050] See also Figure 2 , is a schematic diagram of the structure of a user reputation calculation device based on time series provided in an embodiment of the present application. The user reputation calculation device based on time series can be implemented as all or part of the device through software, hardware or a combination of both. The device includes a data acquisition module 11, a first weight module 12, a second weight module 13, a reputation scoring module 14 and a scoring adjustment module 15.

[0051] A data acquisition module 11 is used to acquire behavior data of each target user in at least one target social platform; A first weight module 12 is used to determine platform behavior feature data of a single target social platform based on the behavior data, and to perform cluster analysis on each target social platform based on the platform behavior feature data to obtain a first weight corresponding to each target social platform; The second weight module 13 is used to determine the individual behavior characteristic data of a single target user on each target social platform according to the behavior data, perform cluster analysis on each target user according to the individual behavior characteristic data, obtain a second weight corresponding to each target user, and determine the behavior weight of a single behavior of the corresponding target user according to the second weight; A reputation scoring module 14, configured to determine an initial reputation score of a corresponding target user based on the behavior data of a single target user on each target social platform, each first weight and each corresponding behavior weight; The score adjustment module 15 is used to obtain at least one future reputation score of a single target user in the target social platform based on a preset time series model, and determine the final reputation score of the corresponding target user according to the initial reputation score and the future reputation score corresponding to the same target user. The time series model is used to predict the reputation score of the target user after the current time based on the reputation score of the target user before the current time. The future reputation score is the instantaneous reputation score within a preset time after the current time.

[0052] Optionally, the first weight module 12 is specifically configured to: Based on the platform behavior feature data corresponding to a single target social platform, construct a platform feature vector of the corresponding target social platform; Performing cluster analysis on the feature vectors of each platform using a preset clustering algorithm to obtain a corresponding first clustering result; According to the first clustering result, all target social platforms are divided into at least one first clustering group, and according to the first clustering group to which each target social platform belongs, a basic weight of the corresponding target social platform is determined; The basic weight corresponding to a single target social platform is adjusted and optimized to obtain an adjusted weight, and each adjusted weight is normalized to obtain a first weight corresponding to each target social platform.

[0053] Optionally, the first weight module 12 is specifically configured to: Determine, based on the platform behavior characteristic data of a single target social platform, a weight adjustment factor of at least one dimension of the corresponding target social platform, where the weight adjustment factor reflects the degree of influence of the platform behavior characteristic data of different dimensions on the basic weight; A weighted sum is performed on each weight adjustment factor to obtain a sum result, and the sum result and the basic weight of the corresponding target social platform are summed to obtain an adjusted weight.

[0054] Optionally, the second weight module 13 is specifically used for: Based on the individual behavior feature data of a single target user, construct the corresponding individual feature vector of the target user; Performing cluster analysis on each individual feature vector using a preset clustering algorithm to obtain a corresponding second clustering result; According to the second clustering result, all target users are divided into at least one second clustering group, and according to the second clustering group to which each target user belongs, a second weight of the corresponding target user is determined.

[0055] Optionally, the second weight module 13 is specifically used for: Determine at least one evaluation index weight for each behavior of the corresponding target user according to the behavior data of the single target user in each target social platform, where the evaluation index weight is an index weight for evaluating the importance of a single behavior of the target user; Sum the weights of each evaluation indicator to obtain the sum of the weights of the corresponding behaviors; The sum of the weights is multiplied by the second weight of the corresponding target user to obtain the behavior weight of each behavior of the corresponding target user.

[0056] Optionally, the reputation scoring module 14 is specifically used for: According to the behavior data of a single target user on each target social platform, determine the behavior frequency of a single behavior of the corresponding target user on a single target social platform; According to each first weight, the behavior frequency and behavior weight corresponding to a single behavior of a single target user in a single target social platform, the initial reputation score of the corresponding target user is determined by a preset reputation score calculation formula. The reputation score calculation formula is: ; In the formula, S represents the initial reputation score, j represents the jth target social platform, M represents the total number of target social platforms, i represents the i-th behavior, and N represents the total number of different types of behaviors of the target user on the target social platform. represents the first weight of the jth target social platform, represents the behavior weight corresponding to the target user's ith behavior, represents the behavior frequency of the target user's i-th behavior in the j-th target social platform, represents the preset error damping factor, represents the preset variance damping factor, represents the error coefficient of the target user's i-th behavior in the j-th target social platform, Represents the variance coefficient of the target user's i-th behavior in the j-th target social platform.

[0057] Optionally, the score adjustment module 15 is specifically used for: Determine a corresponding reputation score error according to the future reputation score of a single target user and the corresponding actual reputation score, where the actual reputation score is the actual value of the reputation score within a preset time after the current time; Normalize each reputation score error to obtain the processed score error, and calculate the target mean and target standard deviation of all processed score errors; Substitute the target mean, target standard deviation and initial reputation score of a single target user into the preset reputation score adjustment formula to obtain the final reputation score of the corresponding target user. The reputation score adjustment formula is: ; In the formula, Z represents the final reputation score of the target user, S represents the initial reputation score of the target user, represents the target mean corresponding to the target user, represents the target standard deviation corresponding to the target user, represents the weight factor of the target mean, A weighting factor representing the target standard deviation.

[0058] It should be noted that the user reputation calculation device based on time series provided in the above embodiment only uses the division of the above functional modules as an example when executing the user reputation calculation method based on time series. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the user reputation calculation device based on time series provided in the above embodiment and the user reputation calculation method based on time series belong to the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.

[0059] The embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, a user reputation calculation method based on time series of the above embodiment is adopted.

[0060] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer-readable medium includes but is not limited to the above-mentioned components.

[0061] Among them, through this computer-readable storage medium, a user reputation calculation method based on time series in the above embodiment is stored in a computer-readable storage medium, and is loaded and executed on a processor to facilitate the storage and application of the above method.

[0062] The embodiment of the present application further discloses an electronic device, wherein a computer program is stored in a computer-readable storage medium, and when the computer program is loaded and executed by a processor, the above-mentioned method for calculating user reputation based on time series is adopted.

[0063] The electronic device may be a desktop computer, a laptop computer, a cloud server or other electronic device, and the electronic device includes but is not limited to a processor and a memory. For example, the electronic device may also include input and output devices, a network access device, and a bus.

[0064] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0065] Among them, the memory can be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device, or it can be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the electronic device. Moreover, the memory can also be a combination of an internal storage unit and an external storage device of the electronic device. The memory is used to store computer programs and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0066] Among them, through this electronic device, a user reputation calculation method based on time series in the above embodiment is stored in the memory of the electronic device, and is loaded and executed on the processor of the electronic device for easy use.

[0067] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. This application is intended to cover any variation, use or adaptive change of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field that are not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for calculating user reputation based on time series, characterized in that: The method comprises: Obtaining behavior data of each target user in at least one target social platform; Determining platform behavior feature data of a single target social platform according to the behavior data, and performing cluster analysis on each target social platform according to the platform behavior feature data to obtain a first weight corresponding to each target social platform; Determine individual behavior feature data of a single target user on each target social platform based on the behavior data, perform cluster analysis on each target user based on the individual behavior feature data to obtain a second weight corresponding to each target user, and determine the behavior weight of a single behavior of the corresponding target user based on the second weight; Determining an initial credibility score of a corresponding target user according to the behavior data of a single target user on each of the target social platforms, each of the first weights and the corresponding behavior weights; Based on a preset time series model, at least one future credibility score of a single target user in the target social platform is obtained, and according to the initial credibility score and the future credibility score corresponding to the same target user, the final credibility score of the corresponding target user is determined. The time series model is used to predict the credibility score of the target user after the current time based on the credibility score of the target user before the current time, and the future credibility score is an instant credibility score within a preset time after the current time.

2. The method for calculating user reputation based on time series according to claim 1, characterized in that: The performing cluster analysis on each of the target social platforms according to the platform behavior characteristic data to obtain a first weight corresponding to each of the target social platforms specifically includes: Based on the platform behavior feature data corresponding to a single target social platform, construct a platform feature vector of the corresponding target social platform; Performing cluster analysis on each of the platform feature vectors using a preset clustering algorithm to obtain a corresponding first clustering result; According to the first clustering result, all the target social platforms are divided into at least one first clustering group, and according to the first clustering group to which each target social platform belongs, a basic weight of the corresponding target social platform is determined; The basic weight corresponding to a single target social platform is adjusted and optimized to obtain an adjusted weight, and each adjusted weight is normalized to obtain a first weight corresponding to each target social platform.

3. The method for calculating user reputation based on time series according to claim 2, characterized in that: The adjusting and optimizing the basic weight corresponding to the single target social platform to obtain the adjusted weight specifically includes: Determine, according to the platform behavior characteristic data of a single target social platform, a weight adjustment factor of at least one dimension of the corresponding target social platform, wherein the weight adjustment factor reflects the degree of influence of the platform behavior characteristic data of different dimensions on the basic weight; A weighted sum is performed on each of the weight adjustment factors to obtain a sum result, and the sum result and the basic weight of the corresponding target social platform are summed to obtain an adjusted weight.

4. The method for calculating user reputation based on time series according to claim 1, characterized in that: The performing cluster analysis on each of the target users according to the individual behavior characteristic data to obtain a second weight corresponding to each target user specifically includes: Based on the individual behavior feature data of a single target user, construct an individual feature vector of the corresponding target user; Performing cluster analysis on each of the individual feature vectors using a preset clustering algorithm to obtain a corresponding second clustering result; According to the second clustering result, all the target users are divided into at least one second clustering group, and according to the second clustering group to which each target user belongs, a second weight of the corresponding target user is determined.

5. The method for calculating user reputation based on time series according to claim 1, characterized in that: Determining the behavior weight of a single behavior of the corresponding target user according to the second weight specifically includes: Determine, according to the behavior data of a single target user in each target social platform, at least one evaluation index weight for each behavior of the corresponding target user, wherein the evaluation index weight is an index weight for evaluating the importance of a single behavior of the target user; The weights of the evaluation indicators are summed to obtain the sum of the weights of the corresponding behaviors; The sum of the weights is multiplied by the second weight of the corresponding target user to obtain the behavior weight of each behavior of the corresponding target user.

6. The method for calculating user reputation based on time series according to claim 1, characterized in that: Determining the initial credibility score of the corresponding target user according to the behavior data of the single target user on each of the target social platforms, each of the first weights and the corresponding behavior weights specifically includes: Determine the frequency of a single behavior of a corresponding target user on a single target social platform according to the behavior data of the single target user on each target social platform; According to each of the first weights, the behavior frequency and behavior weight corresponding to a single behavior of a single target user in a single target social platform, the initial reputation score of the corresponding target user is determined by a preset reputation score calculation formula, and the reputation score calculation formula is: ； In the formula, S represents the initial reputation score, j represents the jth target social platform, M represents the total number of target social platforms, i represents the i-th behavior, and N represents the total number of different types of behaviors of the target user on the target social platform. represents the first weight of the jth target social platform, represents the behavior weight corresponding to the target user's ith behavior, represents the behavior frequency of the target user's i-th behavior in the j-th target social platform, represents the preset error damping factor, represents the preset variance damping factor, represents the error coefficient of the target user's i-th behavior in the j-th target social platform, Represents the variance coefficient of the target user's i-th behavior in the j-th target social platform.

7. The method for calculating user reputation based on time series according to claim 1, characterized in that: The determining of the final reputation score of the corresponding target user according to the initial reputation score and the future reputation score corresponding to the same target user specifically includes: Determine a corresponding reputation score error according to the future reputation score of a single target user and the corresponding actual reputation score, wherein the actual reputation score is an actual value of the reputation score within a preset time after the current time; Normalizing each of the reputation score errors to obtain processed score errors, and calculating a target mean and a target standard deviation of all processed score errors; Substitute the target mean, target standard deviation and initial reputation score corresponding to a single target user into a preset reputation score adjustment formula to obtain the final reputation score of the corresponding target user. The reputation score adjustment formula is: ； In the formula, Z represents the final reputation score of the target user, S represents the initial reputation score of the target user, represents the target mean corresponding to the target user, represents the target standard deviation corresponding to the target user, represents the weight factor of the target mean, A weighting factor representing the target standard deviation.

8. A user reputation calculation device based on time series, characterized in that: include: A data acquisition module (11), used to acquire behavior data of each target user in at least one target social platform; A first weight module (12) is used to determine platform behavior characteristic data of a single target social platform based on the behavior data, and to perform cluster analysis on each target social platform based on the platform behavior characteristic data to obtain a first weight corresponding to each target social platform; A second weight module (13) is used to determine individual behavior feature data of a single target user on each target social platform based on the behavior data, perform cluster analysis on each target user based on the individual behavior feature data, obtain a second weight corresponding to each target user, and determine the behavior weight of a single behavior of the corresponding target user based on the second weight; A reputation scoring module (14), used to determine an initial reputation score of a corresponding target user based on the behavior data of a single target user on each of the target social platforms, each of the first weights and the corresponding behavior weights; The score adjustment module (15) is used to obtain at least one future reputation score of a single target user in the target social platform based on a preset time series model, and determine the final reputation score of the corresponding target user according to the initial reputation score and the future reputation score corresponding to the same target user, wherein the time series model is used to predict the reputation score of the target user after the current time according to the reputation score of the target user before the current time, and the future reputation score is an instantaneous reputation score within a preset time after the current time.

9. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by a processor, the method according to any one of claims 1 to 7 is adopted.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor loads and executes the computer program, the method according to any one of claims 1 to 7 is adopted.

Images