User portrait file gathering method and system convenient for enterprise service

Abstract

The invention relates to the technical field of data processing, in particular to a user portrait file gathering method and system facilitating enterprise services, and the method comprises the steps: obtaining portrait data of a target user to be subjected to file gathering, and user data of multiple dimensions of the target user in a preset time period; determining a change degree of the user data of each dimension in a predetermined time period, wherein the change degree indicates a change trend of the user data in the predetermined time period; calculating the contribution rate of each dimension according to the user data of each dimension; a first variable degree value of each dimension is determined based on the contribution rate and the change degree of each dimension, the first variable degree value indicates the performance ability of the dimension as the specific feature of the target user, and the first variable degree value is inversely proportional to the performance ability of the dimension as the specific feature of the user; and determining specific characteristics of the target user according to the first variable degree value, and performing portrait file gathering on the portrait data by using the specific characteristics.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for aggregating user portraits for facilitating enterprise services. Background Art

[0002] Portrait profiling is the process of classifying, grouping, or aggregating large amounts of portrait data. This process is commonly used in enterprise services. In existing enterprise services, portrait profiling typically uses machine learning to extract features, calculate similarity, and perform clustering to identify and classify portrait data.

[0003] In some scenarios, when machine learning is used for feature extraction, the most representative dimensional information of the user can be obtained as the user's unique features through principal component analysis (PCA), and then similarity calculation and clustering of the portrait data can be performed based on the user's unique features. Among them, when using the PCA algorithm to obtain the most representative dimensional information of the user, the contribution rate of different dimensional information is involved. The higher the order of the dimensional information, the greater the contribution rate of the dimensional information. Finally, several dimensional information with the highest contribution rate after multiple iterations are selected as the user's unique features. However, in the process of dimensionality reduction of the PCA algorithm, the order of the dimensional information is random, and a larger contribution rate is assigned to the dimensional information with a higher order. However, the dimensional information with a higher order and a higher contribution rate may not be representative, which can easily lead to misjudgment of the user's unique features, and then lead to low recognition accuracy of the user's unique features, which ultimately affects the accuracy and practicality of the portrait clustering. Summary of the Invention

[0004] In order to solve the technical problem of low accuracy in identifying unique features of users, the present invention aims to provide a method and system for user portrait aggregation that facilitates enterprise services. The technical solutions adopted are as follows:

[0005] In the first aspect, an embodiment of the present application provides a user portrait profiling method for facilitating enterprise services, which includes: obtaining portrait data of the target user to be profiled, and user data of multiple dimensions of the target user within a predetermined time period; determining the degree of change of the user data of each dimension within the predetermined time period, the degree of change indicating the change trend of the user data within the predetermined time period; calculating the contribution rate of each dimension according to the user data of each dimension; determining the first variable degree value of each dimension based on the contribution rate and degree of change of each dimension, the first variable degree value indicating the expressive ability of the dimension as a unique feature of the target user, and the first variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the user; determining the unique features of the target user according to the first variable degree value, and using the unique features to profile the portrait data.

[0006] Optionally, after determining the first variable degree value of each dimension based on the contribution rate and degree of change of each dimension, the method also includes: determining the first representative degree value of each dimension according to each first variable degree value and the user data of each dimension, the first representative degree value indicating the degree of representativeness of the dimension as a unique feature of the target user, and the first representative degree value is proportional to the degree of representativeness of the dimension as a unique feature of the user; determining the correlation between user data of different dimensions; determining the second representative degree value of each dimension according to the first representative degree value and the correlation, the second representative degree value indicating the possibility of the dimension as a unique feature of the target user, and the second representative degree value is proportional to the possibility of the dimension as a unique feature of the target user; determining the unique features of the target user according to the first variable degree value includes: determining the unique features of the target user according to the second representative degree value.

[0007] Optionally, determining the first representative degree value of each dimension based on each first variable degree value and the user data of each dimension includes: determining the mean and variance of the first variable degree value of each dimension; determining the second variable degree value of each dimension based on the mean and variance of each dimension, the second variable degree value indicating the expressive ability of the dimension as a unique feature of the target user, and the second variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the target user; determining the first difference between the first variable degree values of different target users at the same time point, and the second difference between the user data of the same dimension of different target users; determining the first representative degree value of the target user based on the first difference, the second difference and the second variable degree value.

[0008] Optionally, determining the first representative degree value of the target user based on the first difference, the second difference and the second variable degree value includes: normalizing the first difference and the second difference using the norm function to obtain a normalized value; traversing all target users, accumulating the normalized values of all target users to obtain an accumulated normalized value; calculating the ratio of the accumulated normalized value to the second variable degree value to obtain the first representative degree value.

[0009] Optionally, determining the correlation between user data of different dimensions includes: obtaining the maximum projection variance of the user data of different dimensions in the coordinate axis, and the first axis where the maximum projection variance is located in the coordinate axis; determining the angle between the first axis and the second axis where the dimension is located, the angle indicating the correlation between user data of different dimensions, and the value of the angle is proportional to the correlation.

[0010] Optionally, determining the second representative degree value of each dimension based on the first representative degree value and the correlation degree includes: determining the difference between the included angle and the preset angle; calculating the product of the difference and the maximum projection variance, fixing the current dimension, traversing all other dimensions and obtaining the cumulative value of the product; calculating the product of the cumulative value and the first representative degree value to obtain the second representative degree value of each dimension.

[0011] Optionally, determining the unique feature of the target user according to the second representative degree value includes: determining a dimension whose second representative degree value is greater than a threshold as the unique feature of the target user.

[0012] Optionally, determining the degree of change of user data of each dimension within a predetermined time period includes: obtaining a data baseline of user data of each dimension within the predetermined time period; determining the instantaneous baseline slope of user data corresponding to any time point of each dimension within the predetermined time period based on the data baseline; traversing all time points within the predetermined time period, and determining the time difference between any time point within the predetermined time period and the current time point of each dimension; determining the maximum time difference from the time differences; traversing all time points within the predetermined time period, and using the maximum time difference and the time difference of each time point to perform weighted summation on the instantaneous baseline slope of each time point to obtain a sum value; calculating the ratio of the sum value to all time points within the preset time period, and using the ratio as the degree of change.

[0013] In a second aspect, an embodiment of the present application provides a user portrait archiving system that is convenient for enterprise services, and the user portrait archiving system that is convenient for enterprise services includes: an acquisition module for acquiring portrait data of target users to be archived, and user data of multiple dimensions of target users within a predetermined time period; a determination module for determining the degree of change of user data of each dimension within a predetermined time period, and the degree of change indicates the change trend of user data within the predetermined time period; a calculation module for calculating the contribution rate of each dimension based on the user data of each dimension; the determination module is also used to determine the first variable degree value of each dimension based on the contribution rate and degree of change of each dimension, the first variable degree value indicates the expressive ability of the dimension as a unique feature of the target user, and the first variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the user; the determination module is also used to determine the unique features of the target user based on the first variable degree value, and use the unique features to perform portrait archiving on the portrait data.

[0014] In a third aspect, an embodiment of the present application provides a user portrait archiving system for facilitating enterprise services, which includes: a processor and a memory; wherein the memory is used to store computer programs that can be run on the processor; and the processor is used to execute the programs stored in the memory to implement the steps of the user portrait archiving method for facilitating enterprise services as mentioned in the first aspect above.

[0015] The present invention has the following beneficial effects: by obtaining the portrait data of the target user to be profiled, and the user data of multiple dimensions of the target user within a predetermined time period, and then determining the degree of change of the user data of each dimension within the predetermined time period, the degree of change indicates the change trend of the user data within the predetermined time period, and then calculating the contribution rate of each dimension according to the user data of each dimension, and finally determining the first variable degree value of each dimension based on the contribution rate and the degree of change of each dimension, the first variable degree value indicates the expressive ability of the dimension as a unique feature of the target user, the first variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the user, and finally determining the unique features of the target user according to the first variable degree value, and using the unique features to perform portrait profiling on the portrait data. In this way, since user data of different dimensions will change to different degrees over time, for user data of any dimension, the greater the degree of change, the smaller the ability to express the user's unique characteristics. Therefore, the contribution rate and variability of the dimension are combined to determine the user's unique characteristics, avoiding the misjudgment of the user's unique characteristics caused by directly using the contribution rate to obtain the user's unique characteristics, improving the recognition accuracy of the user's unique characteristics, and thus improving the accuracy and practicality of portrait aggregation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A flowchart of a user portrait aggregation method for facilitating enterprise services provided by one embodiment of the present invention;

[0018] Figure 2 A flowchart of another method for gathering user portraits for enterprise services provided by one embodiment of the present invention;

[0019] Figure 3 A schematic diagram of the module composition of a user portrait archiving system for enterprise services provided by one embodiment of the present invention;

[0020] Figure 4 A schematic structural diagram of a user portrait archiving system for facilitating enterprise services provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation methods, structures, features and effects of the user portrait archiving method and system proposed by the present invention for facilitating enterprise services. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics of one or more embodiments may be combined in any suitable form.

[0022] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0023] In some scenarios, when machine learning is used for feature extraction, the most representative dimensional information of the user can be obtained as the user-specific feature through principal component analysis (PCA), and then the similarity calculation and clustering of the portrait data can be performed based on the user-specific feature. Among them, when the PCA algorithm is used to obtain the most representative dimensional information of the user, the contribution rate of different dimensional information is involved. The higher the order of the dimensional information, the greater the contribution rate of the dimensional information. Finally, several dimensional information with the highest contribution rate after multiple iterations are selected as the user-specific feature. However, in the process of dimensionality reduction, the order of the dimensional information of the PCA algorithm is random, and a larger contribution rate is assigned to the dimensional information with a higher order. However, the dimensional information with a higher order and a higher contribution rate may not be representative, which can easily lead to misjudgment of the user-specific feature, and then lead to low recognition accuracy of the user-specific feature, thereby affecting the accuracy and practicality of the portrait cluster.

[0024] According to analysis, different dimensional information will change to varying degrees over time. For example, data such as user name and height will change less, while basic data such as weight will fluctuate more. Moreover, for any dimensional information, the greater the fluctuation, the less expressive it is as a unique feature of the user. Therefore, directly using the contribution rate to obtain the sorting priority of the current dimension can easily lead to misjudgment of the user's unique features.

[0025] Secondly, the changes in dimensional information of different dimensions themselves are too limited for users, and for the dimensional information of the same dimension of different users, the single dimensional information itself has certain commonalities and differences. The commonalities are reflected in the similar variability of certain dimensional information, while the differences are reflected in the difference in the numerical values of the different dimensional information themselves. Therefore, only analyzing the ability of a single user's dimensional information in this dimension to express personal characteristics cannot accurately express the user's most representative characteristics, which in turn causes the practicality of the entire aggregation method to be too low.

[0026] The specific scheme of the user portrait aggregation method and system for facilitating enterprise services provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0027] Example 1:

[0028] See also Figure 1 , which shows a flow chart of a method for user portrait aggregation for enterprise services provided by one embodiment of the present invention, including:

[0029] Step S101: Acquire the portrait data of the target user to be archived, as well as the user data of multiple dimensions of the target user within a predetermined time period.

[0030] Specifically, the target user's portrait data includes image data captured by the camera in various scenarios, including facial images of the target user, where there can be multiple target users. For each target user, the portrait data also includes attribute fields associated with the target user. Attribute fields represent specific attributes of a data entity. For example, in a database containing personal information, attribute fields may include name, age, gender, and address. These attribute fields are stored together with the facial images.

[0031] Collect historical data of target users within a predetermined time period. This historical data includes user data of multiple dimensions of the target users, such as the user's household registration data, vehicle and driver management data, related person data, work unit data, ticket purchase data, and browsing data.

[0032] The scheduled time period can be customized. For example, the scheduled time period can be defined as seven days, thereby collecting statistics on user data of all dimensions of the target users within the scheduled time period.

[0033] Step S103: determining the degree of change of the user data of each dimension within a predetermined time period.

[0034] The degree of change indicates a change trend of user data within a predetermined time period.

[0035] Specifically, user data in various dimensions will change to a certain extent over time, which is not only reflected in the variance of the user data itself, but also in the fact that for user data in some dimensions, there may be a certain trend of change over time.

[0036] Furthermore, the degree of change may be represented by a slope change of the user data of the dimension within a predetermined time period. In one possible implementation, determining the degree of change of the user data of each dimension within the predetermined time period includes:

[0037] Obtain a data baseline of user data of each dimension within a predetermined time period, determine the instantaneous baseline slope of user data corresponding to any time point within the predetermined time period for each dimension based on the data baseline, traverse all time points within the predetermined time period, and determine the time difference between any time point within the predetermined time period and the current time point for each dimension; determine the maximum time difference from the time difference; traverse all time points within the predetermined time period, traverse all time points within the predetermined time period, and use the maximum time difference and the time difference of each time point to perform weighted summation on the instantaneous baseline slope of each time point to obtain a sum value; calculate the ratio of the sum value to all time points within the preset time period, and use the ratio as the degree of change.

[0038] Specifically, a data baseline refers to the overall trend of user data itself, obtained through iteration. This is similar to obtaining the nonlinear slope of user data over a preset time period. A regression algorithm can be used to obtain the data baseline. After obtaining the data baseline, the instantaneous baseline slope of the user data corresponding to any time point within the preset time period can be obtained using the data baseline as a reference. The time difference between any time point within the preset time period and the current time point is determined for each dimension. This time difference includes the maximum time difference, which is the maximum time difference between any historical time point and the current time point within the preset time period. All time points within the preset time period are traversed, and the weight of the instantaneous baseline slope at each time point is determined using the maximum time difference and the time difference for each time point. This weight can be the difference between 1 and the ratio of the time difference at each time point to the maximum time difference. The instantaneous baseline slope of each time point is then weighted using this difference as the weight. The weighted sum of all time points is calculated and compared with the number of time points. The resulting ratio represents the degree of change.

[0039] Step S105 , calculating the contribution rate of each dimension based on the user data of each dimension.

[0040] Specifically, the contribution rate refers to the variance of user data in each dimension. When calculating the variance of each dimension, we first calculate the average of the user data in each dimension, then calculate the square of the difference between each user data and the average, and then add up the squares of the difference between each user data and the average and compare them with the total user data to obtain the variance of each dimension.

[0041] Step S107 : determining a first variability value of each dimension based on the contribution rate and variation degree of each dimension.

[0042] The first variable degree value indicates the ability of the dimension to be expressed as a unique characteristic of the target user, and the first variable degree value is inversely proportional to the ability of the dimension to be expressed as a unique characteristic of the user.

[0043] Specifically, the product of the contribution rate and the degree of variation can be used as the first variability value. A larger first variability value indicates a higher degree of variability in the user data, but a lower ability to represent the user's personal characteristics, i.e., a lower representativeness of the dimension as a user-specific characteristic.

[0044] Furthermore, the following formula may be used to determine the first variable degree value:

[0045]

[0046] Where i represents the user data of any dimension, r represents any collection time point within the preset time period of the dimension, Δtr Indicates the time difference between any collection time point (current time point) r and the current time point within the preset time period, maxΔt indicates the maximum time difference, k r Indicates the instantaneous baseline slope at any time point within any preset time period; N i Indicates the number of time points within the preset time period, A i Represents the dataset of the current dimension, This value represents the data variance of a dataset within a preset time period. The current time point is the most recent time point corresponding to the collected user data, and the maximum time difference, maxΔt, is the maximum time difference between any time point within the preset time period and the current time point. For example, suppose 10 seconds of user data are collected, the first 9 seconds are used as historical data, and any time point within the first 9 seconds is used as the current time point, i.e., r. The last 10 seconds is used as the current time point.

[0047] Then for the above formula, That is, it represents the time difference between the time series position corresponding to the current time point r and the current time point. The larger the difference, the lower the confidence of the change tendency of the data corresponding to the current time point, and the lower its reference weight. Therefore, Then, using the above The instantaneous baseline slope of each time point is weighted as a weight, and all current time points within the preset time period are traversed to obtain This value has positive and negative values. When the value is greater than 0 and the mean is larger, it means that the user data of the current dimension has a higher feature expression ability for the user. When it is less than 0, it is the opposite. When it is 0, it means that it basically does not change as a whole and belongs to fixed dimension data. Then use the contribution rate of each dimension obtained in the PCA algorithm, that is, the variance, to calculate the above mean. Weighted, we get G i , the value G i The larger the value, the higher the variability of user data and the lower its ability to express user personal characteristics.

[0048] Step S109: determining the unique features of the target user according to the first variable degree value, and performing portrait clustering on the portrait data using the unique features.

[0049] Specifically, the higher the first variability value, the lower the ability to express the user's personal characteristics. Therefore, after normalizing the first variability value in a positive proportion, the dimension whose normalized first variability value is less than a set threshold value can be selected as the target user's unique characteristics, where the set threshold value can be determined based on actual application and is not limited in this embodiment of the present application. After determining the user's unique characteristics, machine learning can be used to calculate similarity and cluster the portrait data based on the user's unique characteristics.

[0050] Through the technical solution disclosed in the embodiments of the present application, by obtaining the portrait data of the target user to be profiled, and the user data of multiple dimensions of the target user within a predetermined time period, the degree of change of the user data of each dimension within the predetermined time period is determined, and the degree of change indicates the change trend of the user data within the predetermined time period. Then, the contribution rate of each dimension is calculated respectively according to the user data of each dimension. Finally, based on the contribution rate and degree of change of each dimension, the first variable degree value of each dimension is determined respectively. The first variable degree value indicates the expressive ability of the dimension as a unique feature of the target user. The first variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the user. Finally, the unique features of the target user are determined according to the first variable degree value, and the portrait data is profiled using the unique features. In this way, since user data of different dimensions will change to different degrees over time, for user data of any dimension, the greater the degree of change, the smaller the ability to express the user's unique characteristics. Therefore, the contribution rate and variability of the dimension are combined to determine the user's unique characteristics, avoiding the misjudgment of the user's unique characteristics caused by directly using the contribution rate to obtain the user's unique characteristics, improving the recognition accuracy of the user's unique characteristics, and thus improving the accuracy and practicality of portrait aggregation.

[0051] Example 2:

[0052] According to the first embodiment above, the variability value of the user data of any task is obtained. At this time, for the user data of the same dimension for different users, the variability is specifically manifested in the fact that the single-dimensional data itself has certain commonalities and differences. The commonality is reflected in the relatively similar variability of the user data of certain dimensions, while the difference is reflected in the difference in the numerical values of the user data of different dimensions. Therefore, only analyzing the ability of the user data of a single user in that dimension to express the user's unique characteristics, and then judging the representativeness of that dimension, may not accurately determine the most representative characteristics of the user. According to the analysis, the variability of the user data of any dimension among all statistical users and the difference in the numerical values of the user data of the current dimension determine the ability of the user data of the current user in the current dimension to express the user's unique characteristics. When the difference in the variability values of different users is smaller, it means that the average tendency of the user data change in the current dimension is smaller. At the same time, when the difference in the variability values of different users is larger, it means that other users do not have the same characteristics as that of the user, and its representativeness is higher, otherwise it is lower. Specifically, if both the mean and the difference in the tendency of user data change for a given dimension are small, it indicates that all users possess that characteristic, making it less representative of the current user's unique characteristics. However, if both are large, it indicates a high degree of representativeness. Therefore, the above logic can be used to determine the degree to which any dimension represents a unique characteristic of the target user.

[0053] See also Figure 2 , which shows a flowchart of another method for gathering user portraits for enterprise services provided by an embodiment of the present invention. The method for gathering user portraits for enterprise services may specifically include the following steps:

[0054] Step S201 : obtaining portrait data of a target user to be profiled, and user data of multiple dimensions of the target user within a predetermined time period.

[0055] Step S203 : determining the degree of change of the user data of each dimension within a predetermined time period, where the degree of change indicates a change trend of the user data within the predetermined time period.

[0056] Step S205 , calculating the contribution rate of each dimension based on the user data of each dimension.

[0057] Step S207: Determine the first variable degree value of each dimension based on the contribution rate and degree of change of each dimension. The first variable degree value indicates the expressive ability of the dimension as a unique feature of the target user. The first variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the user.

[0058] It is worth noting that step S201 and step S207 have the same or similar implementation methods as steps S101 to S107 in the above embodiment, and they can refer to each other, and the embodiments of the present application will not be repeated here.

[0059] Step S2080 : determining a first representative degree value of each dimension according to each first variable degree value and the user data of each dimension.

[0060] The first representative degree value indicates the representative degree of the dimension as a unique feature of the target user, and the first representative degree value is proportional to the representative degree of the dimension as a unique feature of the user.

[0061] Specifically, the first representativeness value refers to the dimension that is most representative for the current user, determined by combining user data from different user dimensions. The first variability is first used to determine the overall variability of the user data for the current dimension relative to the user data for the same dimension for all other users. This overall variability is then used to determine the data differences between the user data for the same dimension for other users. This overall variability and data differences are then used to determine the first representativeness value. A larger first representativeness value indicates a more representative user data for the current user, a higher priority for the user data in the PCA dimensionality reduction, and a higher representativeness of the user data as a unique characteristic.

[0062] In one possible implementation, determining the first representative degree value for each dimension based on each first variable degree value and user data for each dimension includes: determining the mean and variance of the first variable degree value for each dimension; determining the second variable degree value for each dimension based on the mean and variance for each dimension, wherein the second variable degree value indicates the ability of the dimension to represent a unique characteristic of the target user, and the second variable degree value is inversely proportional to the ability of the dimension to represent a unique characteristic of the target user; determining a first difference between the first variable degree values of different target users at the same time point, and a second difference between the user data of different target users for the same dimension; and determining the first representative degree value for the target user based on the first difference, the second difference, and the second variable degree value.

[0063] Specifically, the mean and variance of the first variability value of the user data of the same dimension for each user among all users are calculated respectively, where the mean represents the global universality of the user data of the dimension, and the variance represents the authenticity of the user data of the dimension. The second variability value can be calculated using the following formula:

[0064]

[0065] For the above formula, That is, it represents the average of the first variability values of the current dimension i among all users. The larger the value, the greater the tendency of the user data of this dimension to change, and the lower the ability to express the user's personal characteristics. A set of numbers representing the first variable degree values of all users G The variance of the dimension is larger, the larger the value is, the higher the degree of fluctuation of the user data of different users is, and the lower the authenticity of the overall variability of the dimension is. Therefore, it is expressed by fraction, and thus μ is obtained. i , that is, the overall variability of the current dimension data after weighting, that is, the second variability value.

[0066] Next, for the current user, the difference between the user data of the current user and other users corresponding to the same dimension within a predetermined time period is determined (i.e., the second difference), and in this process, the difference between the first variable degree values of the same dimension of different users (i.e., the first difference) is used as the weight of the difference between the user data of the current user and other users corresponding to the same dimension within a predetermined time period, and the first representative degree value is determined by weighted summation. The first representative degree value can indicate the representative degree of the current dimension of the current user as a unique feature of the user.

[0067] In one possible implementation, determining the first representative degree value of the target user based on the first difference, the second difference and the second variable degree value includes: normalizing the first difference and the second difference using the norm function to obtain a normalized value; traversing all target users, accumulating the normalized values of all target users to obtain an accumulated normalized value; calculating the ratio of the accumulated normalized value to the second variable degree value to obtain the first representative degree value.

[0068] Specifically, the larger the second difference is, the greater the difference in dimensional features between the two users is, and the more representative the user data of the current dimension is for the current user features. At this time, the difference in the degree of variability of the user data of the dimension corresponding to the two users is used as a penalty item to weight the difference in dimensional features according to the above logic. At this time, the greater the difference in the degree of variability is, the higher the representativeness is, and vice versa. Therefore, the norm function is used to normalize the first difference and the second difference.

[0069] More specifically, the first representative degree value can be calculated using the following formula:

[0070]

[0071] Where: x represents any other user, 0 represents the current user, T 0,r T x,r That is, it represents the user data of the current dimension corresponding to different users at the same time, G 0,i G x,i That is, it indicates the variability of different users' corresponding user data at the same time, and M represents the total number of other users.

[0072] Then for this formula: T 0,r -T x,r Indicates the difference in user data of the current dimension corresponding to different users at the same time, traversing all time points within the preset time period, and thus obtaining The larger the value, the greater the difference in dimensional features between the two users, and the higher the degree of representation of the user data of the current dimension for the current user features; at this time, the difference in the degree of variability of the user data of the dimension corresponding to the two users is used as a penalty item to weight the numerical difference according to the above logic. At this time, the greater the difference in the degree of variability, the higher the representativeness, and vice versa. Therefore, the norm function is used for normalization. At this time, by traversing all users, we can get The larger the value is, the greater the overall variability μ of the current dimension itself is. i The size of is expressed by fraction according to its inverse relationship, which is used as a penalty term and weighted to obtain S iThe larger the value, the more representative the user data of the current dimension is for the current user, and the higher the dimensionality reduction priority of the dimension data in PCA dimensionality reduction.

[0073] Step S2081: Determine the correlation between user data of different dimensions.

[0074] Specifically, the correlation degree refers to the correlation between the user data of the current dimension and the user data of other dimensions. The higher the correlation degree, the higher the possibility of it being an independent feature. The correlation degree can be determined by the angle between the corresponding axes of the user data of different dimensions in the coordinate axis.

[0075] In one possible implementation, obtaining the correlation between user data of different dimensions includes: obtaining the maximum projection variance of the user data of different dimensions in the coordinate axis, and the first axis where the maximum projection variance is located in the coordinate axis; determining the angle between the first axis and the second axis where the dimension is located, the angle indicating the correlation between the user data of different dimensions, and the value of the angle is proportional to the correlation.

[0076] Specifically, the PCA algorithm is used to represent the centralized user data of different dimensions on the same coordinate axis. The maximum projected variance between the user data of two different dimensions and the first axis on which the maximum projected variance lies in the coordinate axis are obtained. The smaller the angle between the first axis and the second axis on which the current dimension lies, the smaller the correlation between the user data of the current dimension and the user data of other dimensions, and the higher the possibility of it being an independent feature. At the same time, the larger the maximum projected variance, the higher the ability of the new dimension after dimensionality reduction to represent the characteristics of the user data of the two dimensions.

[0077] Step S2082: determining the second representative degree value of each dimension according to the first representative degree value and the correlation degree.

[0078] The second representative degree value indicates the possibility that the dimension is a unique feature of the target user, and the second representative degree value is proportional to the possibility that the dimension is a unique feature of the target user.

[0079] Specifically, the second representative degree value shows the representativeness of the user data of the current dimension for the current user. The larger the second representative degree value, the more representative the user data of the current dimension is as the user's unique characteristics, and the higher the possibility that it is the most representative unique characteristics of the user.

[0080] In one possible implementation, determining the second representative degree value of each dimension based on the first representative degree value and the correlation degree includes: determining the difference between the included angle and the preset angle; calculating the product of the difference and the maximum projection variance, fixing the current dimension, traversing all other dimensions and obtaining the cumulative value of the product; calculating the product of the cumulative value and the first representative degree value to obtain the second representative degree value of each dimension.

[0081] Specifically, the preset angle may be 90 degrees, and the second representative degree value may be calculated using the following formula:

[0082]

[0083] Where: j represents any other dimension, n represents the total number of other dimensions, θ i,j Indicates the angle between the first axis where the maximum projection variance is located and the second axis where the current dimension i is located, maxσ i,j That is, it represents the maximum projection variance between the current dimension i and other dimensions j, S i That is, it indicates the representativeness of the current dimension of the current user itself, that is, the first representative degree value.

[0084] Then for the above formula, Represents the difference between the angle and 90 degrees. The larger the difference, the smaller the angle between the axis where the maximum projection variance is located and the axis where the current dimension is located. According to logic, the higher the possibility of it being an independent feature. At the same time, the larger the maximum projection variance, the higher the ability of the new dimension after the dimensionality reduction of the two dimensions to express the characteristics of the user data of the two dimensions. Therefore, the two are multiplied, the current dimension is fixed, all dimensions are traversed and the cumulative value is obtained, thus obtaining At the same time, combined with the first representative degree value in the above step, the former is weighted to obtain Y, which represents the representativeness of the user data of the current dimension for the current user. The larger the value, the more representative it is as a unique feature of the user, and the higher the possibility that it is the most representative unique feature of the user.

[0085] Step S209 : determining the unique features of the target user according to the second representative degree value, and clustering the portrait data using the unique features.

[0086] Specifically, the second representative degree value shows the representativeness of the user data of the current dimension for the current user. The larger the value, the higher its representativeness as the user's unique feature, and the higher the possibility of it being the most representative unique feature of the user. Therefore, the dimension with a second representative degree value greater than a threshold value can be selected as the unique feature of the target user, wherein the threshold value can be set according to actual conditions, and the embodiment of the present application is not limited here. Furthermore, the first variable degree value can be normalized in proportion, and the dimension with a normalized second representative degree value greater than a predetermined threshold value can be selected as the unique feature of the target user. The predetermined threshold value can be set according to actual conditions, and the embodiment of the present application is not limited here, such as the predetermined threshold value is set to 0.7. After determining the unique features of the user, machine learning can be used to perform similarity calculation and clustering on the portrait data based on the unique features of the user.

[0087] Through the technical solution disclosed in the embodiments of the present application, since user data of different dimensions will change to different degrees over time, for user data of any dimension, the greater the degree of change, the smaller the ability to express the user's unique characteristics. Therefore, the contribution rate and degree of variability of the dimension are combined to determine the user's unique characteristics, avoiding the misjudgment of the user's unique characteristics caused by directly using the contribution rate to obtain the user's unique characteristics, improving the recognition accuracy of the user's unique characteristics, and thereby improving the accuracy and practicality of portrait aggregation. Furthermore, by analyzing the overall changes in the degree of variability among all users, combining the differences in the degree of variability of user data of the same dimension and the numerical values of user data of the same dimension between different users, the representativeness of the current dimension for the current user is obtained, and then the actual representativeness of the user data of any dimension is judged in combination with the correlation between user data of different dimensions. This operation avoids only analyzing the ability of a single user's user data in this dimension to express personal characteristics by combining the user's data characteristics with the actual characteristics determined between different users, and is unable to accurately and uniquely express the user's most representative characteristics, further improving the recognition accuracy of the user's unique characteristics, selecting the most representative user's unique characteristics, and thereby improving the accuracy and practicality of portrait aggregation.

[0088] Example 3:

[0089] Corresponding to the user portrait archiving method for facilitating enterprise services provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides a user portrait archiving system for facilitating enterprise services. Figure 3 Schematic diagram of the module composition of the user portrait archiving system for enterprise services provided by the embodiment of the present application. The user portrait archiving system for enterprise services is used to perform Figure 1 The user portrait aggregation method described is convenient for enterprise services, such as Figure 3As shown, the user portrait archiving system 300 for facilitating enterprise services includes: an acquisition module 301, used to acquire portrait data of the target user to be archived, and user data of multiple dimensions of the target user within a predetermined time period; a determination module 302, used to determine the degree of change of the user data of each dimension within a predetermined time period, the degree of change indicating the change trend of the user data within the predetermined time period; a calculation module 303, used to calculate the contribution rate of each dimension based on the user data of each dimension; the determination module 302 is also used to determine the first variable degree value of each dimension based on the contribution rate and degree of change of each dimension, the first variable degree value indicating the expressive ability of the dimension as a unique feature of the target user, and the first variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the user; the determination module 302 is also used to determine the unique features of the target user based on the first variable degree value, and use the unique features to perform portrait archiving on the portrait data.

[0090] Through the technical solution disclosed in the embodiments of the present application, since user data of different dimensions will change to different degrees over time, for user data of any dimension, the greater the degree of change, the smaller the ability to express the user's unique characteristics. Therefore, the contribution rate and degree of variability of the dimension are combined to determine the user's unique characteristics, avoiding the misjudgment of the user's unique characteristics caused by directly using the contribution rate to obtain the user's unique characteristics, improving the recognition accuracy of the user's unique characteristics, and thereby improving the accuracy and practicality of portrait aggregation.

[0091] In one possible implementation, the determination module 302 is further used to determine a first representative degree value for each dimension based on each first variable degree value and the user data of each dimension, the first representative degree value indicating the degree of representativeness of the dimension as a unique feature of the target user, and the first representative degree value is proportional to the degree of representativeness of the dimension as a unique feature of the user; determine the correlation between user data of different dimensions; determine a second representative degree value for each dimension based on the first representative degree value and the correlation, the second representative degree value indicating the possibility of the dimension as a unique feature of the target user, and the second representative degree value is proportional to the possibility of the dimension as a unique feature of the target user; and determine the unique feature of the target user based on the second representative degree value.

[0092] In one possible implementation, the determination module 302 is further used to determine the mean and variance of the first variable degree value of each dimension; determine the second variable degree value of each dimension based on the mean and variance of each dimension, the second variable degree value indicates the expressive ability of the dimension as a unique feature of the target user, and the second variable degree value is inversely proportional to the expressive ability of the dimension as a unique feature of the target user; determine the first difference between the first variable degree values of different target users at the same time point, and the second difference between the user data of the same dimension of different target users; determine the first representative degree value of the target user based on the first difference, the second difference, and the second variable degree value.

[0093] In one possible implementation, the determination module 302 is further used to normalize the first difference and the second difference using a norm function to obtain a normalized value; traverse all target users, accumulate the normalized values of all target users, and obtain an accumulated normalized value; calculate the ratio of the accumulated normalized value to the second variable degree value to obtain a first representative degree value.

[0094] In one possible implementation, the acquisition module 301 is further used to obtain the maximum projection variance of user data of different dimensions in the coordinate axis, and the first axis where the maximum projection variance is located in the coordinate axis; the determination module 302 is further used to determine the angle between the first axis and the second axis where the dimension is located, where the angle indicates the correlation between user data of different dimensions, and the value of the angle is proportional to the correlation.

[0095] In one possible implementation, the determination module 302 is further used to determine the difference between the included angle and the preset angle; calculate the product of the difference and the maximum projection variance, fix the current dimension, traverse all other dimensions and obtain the cumulative value of the product; calculate the product of the cumulative value and the first representative degree value to obtain the second representative degree value of each dimension.

[0096] In a possible implementation, the determination module 302 is further configured to determine a dimension whose second representative degree value is greater than a threshold as a unique feature of the target user.

[0097] In one possible implementation, the determination module 302 is further used to obtain a data baseline of user data of each dimension within a predetermined time period; determine the instantaneous baseline slope of the user data corresponding to any time point in each dimension within the predetermined time period based on the data baseline; traverse all time points within the predetermined time period to determine the time difference between any time point in each dimension within the predetermined time period and the current time point; determine the maximum time difference from the time differences; traverse all time points within the predetermined time period, and use the maximum time difference and the time difference of each time point to perform weighted summation on the instantaneous baseline slope of each time point to obtain a sum value; calculate the ratio of the sum value to all time points in the preset time period, and use the ratio as the degree of change.

[0098] It should be noted that the user portrait archiving system for facilitating enterprise services provided in the embodiment of the present application and the user portrait archiving method for facilitating enterprise services provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned user portrait archiving method for facilitating enterprise services, and has the same or similar beneficial effects, and the repetitions will not be repeated.

[0099] Example 4:

[0100] Corresponding to the user portrait archiving method for facilitating enterprise services provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides a user portrait archiving system for facilitating enterprise services, which is used to execute the user portrait archiving method for facilitating enterprise services. Figure 4 A schematic diagram of a user portrait archiving system for enterprise services is provided to implement various embodiments of the present application. Figure 4 The user portrait archiving system for enterprise services may have relatively large differences due to different configurations or performances, and may include one or more processors 401 and memory 402, the memory 402 is used to store computer programs that can be run on the processor, and the processor is used to execute the program stored in the memory to achieve the above Figure 1 or Figure 2 The various steps in the method embodiment are as follows. Memory 402 may be a temporary storage or a persistent storage. The application stored in memory 402 may include one or more modules (not shown), each of which may include a series of computer-executable instructions in a user portrait archiving system that facilitates enterprise services.

[0101] Furthermore, the processor 401 can be configured to communicate with the memory 402 to execute a series of computer-executable instructions in the memory 402 on the user portrait archiving system for facilitating enterprise services. The user portrait archiving system for facilitating enterprise services can also include one or more power supplies 403, one or more wired or wireless network interfaces 404, one or more input / output interfaces 405, and one or more keyboards 406.

[0102] Specifically in this embodiment, the user portrait archiving system for enterprise services includes a processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other through the bus; the memory is used to store computer programs; the processor is used to execute the programs stored in the memory to achieve the above Figure 1 or Figure 2The various steps in the method embodiment are similar to those in the method embodiment, and have the beneficial effects of the above method embodiments. To avoid repetition, the embodiments of this application will not be described again here.

[0103] It should be noted that the user portrait archiving system for facilitating enterprise services provided in the embodiment of the present application and the user portrait archiving method for facilitating enterprise services provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned user portrait archiving method for facilitating enterprise services, and has the same or similar beneficial effects, and the repetitions will not be repeated.

[0104] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A user portrait archiving method for enterprise services, characterized in that: The portrait gathering method comprises: Acquire portrait data of a target user to be profiled, as well as user data of multiple dimensions of the target user within a predetermined time period; determining a degree of change of the user data of each dimension within the predetermined time period, the degree of change indicating a change trend of the user data within the predetermined time period; Calculate the contribution rate of each dimension based on the user data of each dimension; Determining a first variability value for each dimension based on the contribution rate and the degree of variation of each dimension, wherein the first variability value indicates the ability of the dimension to serve as a unique characteristic of the target user, and the first variability value is inversely proportional to the ability of the dimension to serve as a unique characteristic of the user; The unique features of the target user are determined according to the first variable degree value, and the portrait data is subjected to portrait clustering using the unique features.

2. The user portrait aggregation method for facilitating enterprise services according to claim 1, characterized in that: After determining the first variable degree value of each dimension based on the contribution rate of each dimension and the degree of change, the method further includes: Determining a first representative degree value for each dimension based on each first variable degree value and the user data for each dimension, wherein the first representative degree value indicates the degree to which the dimension represents a unique characteristic of the target user, and the first representative degree value is proportional to the degree to which the dimension represents a unique characteristic of the user; Determine the correlation between user data of different dimensions; determining a second representative degree value for each dimension based on the first representative degree value and the association degree, wherein the second representative degree value indicates the likelihood that the dimension is a unique feature of the target user, and the second representative degree value is proportional to the likelihood that the dimension is a unique feature of the target user; The determining of the unique characteristics of the target user according to the first variable degree value includes: The unique features of the target user are determined according to the second representativeness value.

3. The user portrait aggregation method for facilitating enterprise services according to claim 2, characterized in that: Determining the first representative degree value of each dimension according to each first variable degree value and the user data of each dimension includes: determining a mean and a variance of the first variability value for each of the dimensions; determining a second variability value for each dimension based on the mean and variance of each dimension, wherein the second variability value indicates the ability of the dimension to serve as a unique characteristic of the target user, and the second variability value is inversely proportional to the ability of the dimension to serve as a unique characteristic of the target user; respectively determining a first difference between first variability values of different target users at the same time point, and a second difference between user data of the same dimension of different target users; A first representative degree value of the target user is determined according to the first difference value, the second difference value and the second variable degree value.

4. The user portrait aggregation method for facilitating enterprise services according to claim 3, characterized in that: The determining the first representative degree value of the target user according to the first difference, the second difference and the second variable degree value comprises: Normalizing the first difference and the second difference using a norm function to obtain a normalized value; Traverse all target users, accumulate the normalized values of all target users, and obtain the accumulated normalized value; The ratio of the accumulated normalized value to the second variable degree value is calculated to obtain the first representative degree value.

5. The user portrait aggregation method for facilitating enterprise services according to claim 2, characterized in that: Determining the correlation between user data of different dimensions includes: Obtaining the maximum projection variance of user data of different dimensions in the coordinate axis, and the first axis line where the maximum projection variance is located in the coordinate axis; An angle between the first axis and a second axis where the dimension is located is determined, where the angle indicates a degree of association between user data of different dimensions, and a value of the angle is proportional to the degree of association.

6. The user portrait aggregation method for facilitating enterprise services according to claim 5, characterized in that: Determining the second representative degree value of each dimension according to the first representative degree value and the association degree includes: determining a difference between the included angle and a preset angle; Calculate the product of the difference and the maximum projection variance, fix the current dimension, traverse all other dimensions and calculate the accumulated value of the product; The product of the accumulated value and the first representative degree value is calculated to obtain a second representative degree value of each dimension.

7. The user portrait aggregation method for facilitating enterprise services according to claim 2, characterized in that: The determining of the unique characteristics of the target user according to the second representativeness value includes: The dimension whose second representative degree value is greater than a threshold is determined as a unique feature of the target user.

8. The user portrait aggregation method for facilitating enterprise services according to any one of claims 1 to 7, characterized in that: Determining the degree of change of the user data of each dimension within the predetermined time period includes: Obtaining a data baseline of user data of each dimension within the predetermined time period; Determining, based on the data baseline, the instantaneous baseline slope of the user data corresponding to any time point in each dimension within the predetermined time period; Traversing all time points within the predetermined time period, and determining the time difference between any time point within the predetermined time period and the current time point for each dimension; determining a maximum time difference from the time differences; Traversing all time points within the predetermined time period, performing weighted summation on the instantaneous baseline slope of each time point using the maximum time difference and the time difference of each time point to obtain a sum value; The ratio of the sum value to all time points within the preset time period is calculated, and the ratio is used as the degree of change.

9. A user portrait archiving system that facilitates enterprise services, characterized by: include: An acquisition module, configured to acquire portrait data of a target user to be profiled, and user data of multiple dimensions of the target user within a predetermined time period; A determination module, configured to determine a degree of change of the user data of each dimension within the predetermined time period, wherein the degree of change indicates a trend of change of the user data within the predetermined time period; A calculation module, configured to calculate the contribution rate of each dimension based on the user data of each dimension; The determining module is further configured to determine a first variability value for each dimension based on the contribution rate and the degree of variation of each dimension, wherein the first variability value indicates the ability of the dimension to represent a unique characteristic of the target user, and the first variability value is inversely proportional to the ability of the dimension to represent a unique characteristic of the user; The determination module is further configured to determine the unique features of the target user according to the first variable degree value, and perform portrait clustering on the portrait data using the unique features.

10. A user portrait archiving system that facilitates enterprise services, characterized by: include: A processor and a memory; wherein the memory is used to store a computer program that can be run on the processor; A processor is used to execute the program stored in the memory to implement the steps of the user portrait archiving method for facilitating enterprise services as described in any one of claims 1 to 8.

Images