Internet user portrait generation method based on big data

Abstract

The invention relates to the technical field of data analysis, in particular to an internet user portrait generation method based on big data, which comprises the following steps: acquiring search times and search depth of a target data type obtained by browsing a webpage by a user within a monitoring time period, and obtaining search progressive performance of the target data type, obtaining the browsing concentration degree of the target data type in combination with the interspersed search conditions of other data types in the search process of the target data type; based on the change trend of the browsing concentration degree of the target data type in the plurality of continuous monitoring time periods, obtaining the increase degree of the browsing concentration degree; according to the increase degree of the browsing concentration degree, in combination with the browsing concentration degree of the target data type and the recent search time interval of the target data type, the portrait importance degree of each target data type of the user is correctly mined, and when the user is portrayed according to the portrait importance degree of each target data type, the user experience is improved. The accuracy of the user portrait can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a method for generating Internet user portraits based on big data. Background Art

[0002] Big data-based internet user profiles can accurately identify user needs, preferences, and behavioral patterns, helping businesses optimize product design, personalized recommendations, and targeted marketing, significantly improving user experience and business conversion rates. Furthermore, by mining data for potential risks (such as fraud and unusual behavior), platform security and compliance can be enhanced. In the fiercely competitive digital ecosystem, user profiles are a core tool for data-driven decision-making, cost reduction, and efficiency improvement. They are also the foundation for building intelligent services, providing two-way value to both businesses and users.

[0003] Traditional methods use information such as the number of times users browse different types of data or the duration of their browsing to determine the importance of each type of data, thereby creating a user profile. However, this approach fails to consider the impact of frequent changes in different types of data on the importance of a user's profile. Consequently, it fails to accurately determine the importance of a user's profile, affecting the accuracy of the user profile. Summary of the Invention

[0004] In order to solve the technical problem that existing user profiling methods affect the accuracy of the portrait, the purpose of the present invention is to provide a method for generating Internet user portraits based on big data. The technical solutions adopted are as follows: The present invention provides a method for generating an Internet user portrait based on big data, comprising: Obtain the number of searches and search depths for the target data type obtained by browsing web pages during the user monitoring period, and obtain the progressive search performance of the target data type; Obtaining browsing focus on the target data type based on the search progression performance and in combination with the interspersed search conditions of other data types during the search for the target data type; Obtaining a degree of increase in browsing concentration based on a change trend of browsing concentration of the target data type within multiple consecutive monitoring time periods; According to the increase degree of the browsing concentration, combined with the browsing concentration of the target data type and the most recent search time interval of the target data type, the portrait importance of the target data type is obtained.

[0005] In an exemplary embodiment, the process of obtaining the search progressive representation includes: Obtain the search frequency ratio of the target data type; Obtaining a search depth difference between the search depth and a preset search depth; According to the proportion of the number of times and the difference in the search depth, a search progressive performance of the target data type is obtained, and the search progressive performance is proportional to the proportion of the number of times and inversely proportional to the difference in the search depth.

[0006] In an exemplary embodiment, the process of obtaining the interspersed search situation includes: Obtain the number of data types of other data types interspersed between any two adjacent search processes for the target data type, as well as the interspersed search duration; Obtaining a concentration coefficient corresponding to any two search processes according to the number of data types and the interspersed search duration; the concentration coefficient is inversely proportional to the number of types and the interspersed search duration; The concentration coefficients corresponding to any two search processes are combined to obtain the comprehensive concentration coefficient.

[0007] In an exemplary embodiment, the process of acquiring the browsing focus of the target data type includes: The browsing concentration is obtained by integrating the comprehensive concentration coefficient and the search progressive performance.

[0008] In an exemplary embodiment, the process of obtaining the degree of increase in browsing concentration includes: Obtaining a ratio of the number of monitoring time periods in a first monitoring time period, where the first monitoring time period refers to a monitoring time period having a greater browsing concentration on the target data type than an adjacent previous monitoring time period; Obtaining an average value of the difference between the browsing concentration of the target data type in all first monitoring time periods and the previous adjacent monitoring time period; The degree of increase in browsing concentration is obtained according to the proportion of the number of monitoring time periods and the average value of the difference. The degree of increase in browsing concentration is proportional to the proportion of the number of monitoring time periods and the average value of the difference.

[0009] In an exemplary embodiment, the monitoring time period includes a plurality of historical monitoring time periods and a current monitoring time period arranged in chronological order, the most recent search time interval of the target data type is the time interval between the end time of the search for the target data type in the target historical monitoring time period and the end time of the current monitoring time period; the target historical monitoring time period is the previous monitoring time period adjacent to the current monitoring time period; The importance of the portrait is directly proportional to the increase in the browsing concentration and the browsing concentration of the target data type in the current monitoring time period, and is inversely proportional to the most recent search time interval.

[0010] In an exemplary embodiment, the monitoring time period includes a plurality of historical monitoring time periods and a current monitoring time period arranged in chronological order; The process of obtaining the target data type includes: Based on the browsing focus, the initial data type is divided into a focused data type and a non-focused data type; the initial data type includes the target data type, and the target data type includes the focused data type; Obtain the difference in browsing concentration of the focused data type before and after excluding each non-focused data type from the current monitoring time period; According to the interspersed search situation of each non-focused data type during the focused data type search process in the current monitoring time period, combined with the difference in the browsing focus change, the degree of attention diversion of each non-focused data type to the focused data type is obtained; Obtaining a new consideration level for a portrait of each non-focused data type based on the attention shift level and browsing concentration level of each non-focused data type; The non-focused data type corresponding to the new portrait consideration level that is greater than the preset new portrait consideration level threshold is used as the target data type.

[0011] In an exemplary embodiment, the process of acquiring the degree of attention shift includes: Get the percentage of searches for each non-focused data type during the current monitoring period. Get the search duration ratio of each non-focused data type during the focused data type search in the current monitoring time period; Based on the proportion of search times, the proportion of search duration and the difference in browsing concentration changes, the degree of attention diversion of each non-focused data type to the focused data type is obtained; the degree of attention diversion is proportional to the proportion of search times, the proportion of search duration and the difference in browsing concentration changes.

[0012] In an exemplary embodiment, dividing the initial data type into a focused data type and a non-focused data type based on the browsing focus includes: Comparing the browsing concentration of each initial data type during the current monitoring period with a preset browsing concentration threshold; An initial data type corresponding to a browsing concentration greater than the preset browsing concentration threshold is determined as the concentration data type.

[0013] In an exemplary embodiment, after obtaining the portrait importance of the target data type, the big data-based Internet user portrait generation method further includes: visualizing the user ID, user data of the target data type, and the portrait importance.

[0014] The present invention has the following beneficial effects: the present invention combines the number of searches and the search depth of the target data type obtained by the user browsing the web page during the monitoring time period, as well as the interspersed searches of other data types in the search process for the target data type, to obtain the browsing concentration of the target data type, and thus obtain the degree of increase in browsing concentration based on the dynamic changes in browsing concentration, and finally obtain the portrait importance of the target data type based on the degree of increase in browsing concentration, combined with the browsing concentration of the target data type and the most recent search time interval of the target data type. The portrait importance of each target data type is closely related to the actual data situation of the target data type, so that the portrait importance of each target data type of the user can be correctly excavated. When the user is profiled according to the portrait importance of each target data type, the accuracy of the user portrait can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of a method for generating an Internet user portrait based on big data provided by one embodiment of the present invention; Figure 2 This is a flowchart for obtaining a search progressive representation provided by one embodiment of the present invention; Figure 3 This is a flow chart for obtaining interleaved search conditions provided by one embodiment of the present invention; Figure 4 is a flow chart for obtaining target data type provided by one embodiment of the present invention; Figure 5 is a flow chart of obtaining the degree of attention diversion provided by one embodiment of the present invention; Figure 6 This is a flowchart for obtaining the degree of increase in browsing concentration provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0016] To further illustrate the technical means and effects employed by the present invention to achieve its intended objectives, the following detailed description of the specific embodiments, structures, features, and effects of the present invention is provided in conjunction with the accompanying drawings and preferred embodiments. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this invention pertains. The data and information collected in this application were obtained with full consent and authorization.

[0018] This embodiment provides a method for generating internet user profiles based on big data. This method is applicable in scenarios where big data-based internet user profiles can accurately identify user needs, preferences, and behavior patterns, helping companies optimize product design, personalized recommendations, and precision marketing, significantly improving user experience. Furthermore, by mining potential risks (such as abnormal behavior) through data, platform security and compliance can be enhanced.

[0019] like Figure 1 As shown, this embodiment provides a method for generating an Internet user portrait based on big data, comprising the following steps: Step S1: Obtain the number of searches and search depths of the target data type obtained by browsing web pages during the user monitoring period, and obtain the search progressive performance of the target data type; Step S2: Obtain browsing focus on the target data type based on the search progression performance and the interspersed search of other data types during the search for the target data type; Step S3: obtaining an increase degree of browsing concentration based on a change trend of browsing concentration of the target data type within a plurality of consecutive monitoring time periods; Step S4: According to the increase in the browsing concentration, combined with the browsing concentration of the target data type and the most recent search time interval of the target data type, the importance of the portrait of the target data type is obtained.

[0020] The implementation process of each step is described in detail below with reference to the accompanying drawings.

[0021] Step S1: Obtain the number of searches and search depth of the target data type obtained by browsing web pages during the user monitoring period, and obtain the search progressive performance of the target data type.

[0022] This embodiment processes the relevant information of the data type obtained by a user browsing a web page to obtain the importance of the portrait of the corresponding data type, thereby realizing the portrait operation of the user. Therefore, the user involved in this embodiment is any user. After obtaining the user's authorization, the browsing history (including the URL of the page, access IP, access time, page title, etc.) of the required date range is obtained through the BrowsingHistoryView cross-browser tool and exported in the form of a table. Set the Service Worker to listen to the page closing event to obtain the user's stay time on different pages, and store the stay time corresponding to the data such as the URL of the browsed page.

[0023] This embodiment analyzes browsing data within multiple monitoring time periods, wherein the multiple monitoring time periods include multiple historical monitoring time periods and a current monitoring time period arranged in chronological order. In two adjacent monitoring time periods, the end time of the previous monitoring time period and the start time of the next monitoring time period are two adjacent moments. The duration of each monitoring time period is equal, and the specific duration of each monitoring time period is set according to actual needs, for example, each monitoring time period is 1 week. The end time of the current monitoring time period is set to the current moment. The following description will be made using any monitoring time period as an example.

[0024] If a user browses the web multiple times during the monitoring period, they will be able to obtain multiple data information. According to the preset data type classification mechanism, the data information obtained each time is classified into different data types, thereby obtaining the data type corresponding to each data information obtained. In this way, multiple data types can be obtained. For example, when browsing a shopping app, if you browse a specific brand of mobile phone, the corresponding data type classification method is: specific mobile phone model → mobile phone brand → mobile phone → electronic device, then the corresponding data type is electronic device; if you browse jackets, the corresponding data type classification method is: jackets → tops → clothing, then the corresponding data type is clothing. Therefore, it is possible to obtain information on multiple different types of products, such as different data types for electronic devices and clothing.

[0025] The various data types obtained by the user during the monitoring period are defined as various initial data types. In an exemplary embodiment, the initial data type can be directly used as the target data type required by this embodiment, and the subsequent analysis is directly based on the data information of the initial data type, and the target data type screening process described later is no longer performed. As a better implementation method, the initial data type can also be screened to obtain the target data type, and the target data type is more closely aligned with the user portrait requirements. Then, the initial data type includes the target data type, and the target data type is a data type that is part of the initial data type.

[0026] In one exemplary embodiment, the target data type is obtained by filtering the initial data type. Since the target data type is filtered, it is necessary to obtain the browsing focus of each initial data type. Therefore, it is necessary to first obtain the browsing focus of each initial data type. It should be understood that since the target data type is part of the initial data type, when the browsing focus of each initial data type is obtained, the browsing focus of each target data type is also obtained.

[0027] For any initial data type, the user can obtain multiple data information of the initial data type by browsing the web page during the monitoring time period, that is, the initial data type corresponds to multiple browsing processes, and the data type of the data information obtained each time is the initial data type. For example: the user browses the shopping web page of the electronic device multiple times during the monitoring time period, and obtains a data information each time. Accordingly, the number of searches for the initial data type (that is, the number of views) and the search depth of each search are obtained during the monitoring time period. Among them, the number of searches means: the number of views of the browsed data information of the initial data type is recorded from each browsing, and the number of searches for the initial data type is obtained. The search depth refers to the minimum classification of data obtained each time the browsing is performed. The more specific the minimum classification, the deeper the search depth. For example, the search depth corresponding to the specific model of the mobile phone is deeper than the search depth corresponding to the mobile phone brand. Based on these two pieces of information, the search progressive performance of the initial data type is obtained. In an exemplary embodiment, as Figure 2 As shown, a specific acquisition process of the search progressive expression is given below: Step S1-1: Obtain the search frequency ratio of the initial data type.

[0028] Since multiple initial data types are obtained during the monitoring period, the search count for each initial data type can also be obtained. The sum of the search counts for all obtained initial data types is calculated, and then the ratio of the search count for that initial data type to the sum of the search counts is calculated as the search count ratio for that initial data type.

[0029] Step S1-2: Obtain a search depth difference between the search depth and a preset search depth.

[0030] For ease of calculation, the search depth level corresponding to the initial data type is quantified. The more specific the level corresponding to the search depth, that is, the smaller it is, the larger the value of the search depth is. The value of the search depth is the value of the corresponding level. For example, for the four levels of specific mobile phone model → mobile phone brand → mobile phone → electronic device, the search depth of electronic device is 1, the search depth of mobile phone is 2, the search depth of mobile phone brand is 3, and the search depth of specific mobile phone model is 4.

[0031] By adopting the above process, the search depth obtained by each search of the initial data type is quantized to obtain a quantized value of the search depth obtained by each search.

[0032] In view of the above quantization, a search depth is preset for the initial data type. The preset search depth is the minimum level allowed by the initial data type. For example, for a specific mobile phone model, the quantization value of the preset search depth is 4.

[0033] Then, the search depth difference between each search depth and the preset search depth is obtained, that is, the absolute value of the difference between the quantized value of each search depth and the quantized value of the preset search depth. Then, the average of the absolute values ​​of the difference between the quantized values ​​of all search depths of the initial data type within the monitoring time period and the quantized value of the preset search depth is calculated as the search depth difference of the initial data type within the monitoring time period. The search depth difference of the initial data type represents the gap from the preset maximum search depth level. The smaller the search depth difference, the smaller the gap from the preset maximum search depth level, and the deeper the search depth.

[0034] Step S1-3: According to the frequency ratio and the difference in search depth, the search progressive performance of the initial data type is obtained.

[0035] According to the proportion of the number of searches for the initial data type and the difference in search depth of the initial data type, the search progressive performance of the initial data type is obtained.

[0036] The greater the proportion of searches for an initial data type, and the smaller the difference in search depth for that initial data type, the more pronounced the user's search progression for that initial data type, the more likely the user is to favor data of that initial data type, and the more likely it is that they should retain it when creating user profiles. Therefore, search progression is directly proportional to the proportion of searches and inversely proportional to the difference in search depth.

[0037] In an exemplary embodiment, a quantitative method of searching for progressive performance is given as follows: ; in, Represents the search progression of the jth initial data type, represents the difference in search depth of the jth initial data type, exp represents the exponential function with the natural constant e as the base, Express The negative correlation normalization in this embodiment adopts this method. Represents the proportion of searches for the jth initial data type. This calculation formula is essentially a weighted sum of two parameters that affect the search progression performance, with each weight being 0.5.

[0038] Step S2: Based on the search progression performance and combined with the interspersed search conditions of other data types during the search process for the target data type, the browsing focus of the target data type is obtained.

[0039] For the initial data type, the user's search process for the initial data type may be interspersed with searches for other data types. That is, the user may also browse some web pages of other data types during the search process for the initial data type. Therefore, based on the search progression performance of the initial data type and the interspersed searches of other data types during the search process for the initial data type, the browsing focus of the initial data type is obtained.

[0040] In an exemplary embodiment, Figure 3 As shown, a specific acquisition process of the interspersed search situation is given as follows: Step S2-1: Obtain the number of data types of other data types interspersed between any two adjacent search processes for the initial data type, and the interspersed search duration.

[0041] Since the user will search for the initial data type multiple times during the monitoring period, for any two adjacent search processes for the initial data type, the number of data types of other data types interspersed with the search and the interspersed search duration are obtained during the time period between any two adjacent search processes for the initial data type. Among them, the number of data types of other data types searched by the user is obtained during the time period between any two adjacent search processes for the initial data type, thereby obtaining the number of data types of other data types searched by the user. The search duration of each other data type searched by the user during the time period between any two adjacent search processes for the initial data type is obtained, and then the sum of the search durations is added to obtain the interspersed search duration.

[0042] Step S2-2: Obtain the concentration coefficient corresponding to any two search processes based on the number of data types and the interleaved search duration.

[0043] When the number of interleaved searches for other data types is smaller, it means that the user has less interest in other data types during the time period between any two adjacent searches for the initial data type. Therefore, the concentration coefficient corresponding to these two adjacent searches for the initial data type is higher. When the interleaved searches for other data types are shorter, it means that the user has less interest in other data types during the time period between any two adjacent searches for the initial data type. Therefore, the concentration coefficient is inversely proportional to both the number of types and the interleaved search duration.

[0044] In an exemplary embodiment, the product of the number of types and the interspersed search duration is calculated and normalized for negative correlation, and the result obtained is the concentration coefficient corresponding to the two search processes for the initial data type.

[0045] Step S2-3: Fuse all the concentration coefficients corresponding to any two search processes to obtain the comprehensive concentration coefficient of the initial data type.

[0046] Through step S2-2, the concentration coefficient corresponding to any two adjacent search processes for the initial data type is obtained, and the concentration coefficient corresponding to all any two search processes of the initial data type is integrated. Specifically, the average value of the concentration coefficient is calculated, and the result obtained is the comprehensive concentration coefficient of the initial data type.

[0047] The greater the comprehensive focus coefficient, the more meaningful the initial data type is for user profiling, that is, the greater the browsing focus of the initial data type. The greater the user's search progression, the more meaningful the initial data type is for user profiling, that is, the greater the browsing focus of the initial data type.

[0048] The comprehensive concentration coefficient and search progressive performance of the initial data type are integrated to obtain the browsing concentration of the initial data type. The browsing concentration of the initial data type is proportional to the comprehensive concentration coefficient and search progressive performance of the initial data type. In an exemplary embodiment, a specific fusion method is given as follows: the comprehensive concentration coefficient of the initial data type and the search progressive performance of the initial data type are weighted and summed, and the weights are all 0.5, that is, the average value of the comprehensive concentration coefficient of the initial data type and the search progressive performance of the initial data type is calculated, and the result obtained is the browsing concentration of the initial data type.

[0049] Using the above process, we can obtain the browsing focus of each initial data type. Then, based on the browsing focus of each initial data type, we can filter the target data type from each initial data type. To improve the reliability of the target data type screening, the target data type is obtained from the current monitoring time period.

[0050] In an exemplary embodiment, Figure 4 As shown, a specific process of obtaining the target data type is given below: Step S2-4: Based on the browsing concentration, the initial data type is divided into the focused data type and the non-focused data type.

[0051] The above method is used to obtain the browsing concentration of each initial data type in the current monitoring time period. In order to improve the reliability of subsequent screening, after obtaining the browsing concentration of each initial data type in the current monitoring time period, the maximum and minimum values ​​of the browsing concentration are obtained, and then the maximum and minimum normalization method is used to normalize the browsing concentration of each initial data type in the current monitoring time period. The browsing concentration of each initial data type in the current monitoring time period used for subsequent comparisons are all normalized results using the maximum and minimum normalization method.

[0052] A browsing focus threshold is preset. This threshold is used to determine whether the browsing focus of each initial data type is high during the current monitoring period, thereby separating the focused data type from the non-focused data type. The value range of this threshold is 0-1. The specific value of this threshold is set based on actual needs, provided that the comparison is satisfied. In this embodiment, 0.7 is used as an example.

[0053] Compare the browsing concentration of each initial data type in the current monitoring time period with the preset browsing concentration threshold, and determine the initial data type corresponding to the browsing concentration greater than the preset browsing concentration threshold as the focused data type. Accordingly, the other initial data types except the focused data type are non-focused data types, thereby dividing the initial data types into focused data types and non-focused data types. Take the focused data type as the target data type, then the target data type includes the focused data type. The purpose of the subsequent steps is to filter out a part of the data types from the non-focused data type as the target data type, so as to realize the addition and update of the target data type.

[0054] Step S2-5: Obtain the difference in browsing concentration of the focused data type before and after each non-focused data type is removed from the current monitoring time period.

[0055] The following explanation is given using any non-focused data type as an example. Obtain the browsing concentration of the non-focused data type. Then remove the non-focused data type from the current monitoring time period, and the distribution of data types in the current monitoring time period is changed. Since browsing concentration is related to not only its own data type but also other data types in the current monitoring time period, after removing the non-focused data type from the current monitoring time period, the browsing concentration of other data types will change.

[0056] For any focused data type, obtain the browsing focus of the focused data type before the non-focused data type is removed from the current monitoring time period. Also obtain the browsing focus of the focused data type after the non-focused data type is removed from the current monitoring time period. Then, obtain the difference in browsing focus of the focused data type before and after the non-focused data type is removed from the current monitoring time period. The difference in browsing focus of the focused data type is specifically the absolute value of the difference in browsing focus of the focused data type before and after removal.

[0057] The above calculation method is used for each focus data type, thereby obtaining the browsing focus change difference corresponding to each focus data type with respect to the non-focus data type. Then, the average browsing focus change difference is calculated for all focus data types to obtain the browsing focus change difference corresponding to the non-focus data type. The calculation formula is as follows: ; in, Indicates the difference in browsing focus change corresponding to the g-th non-focus data type, It represents the browsing concentration of the i-th focused data type before the g-th non-focused data type is removed from the current monitoring period. It represents the browsing concentration of the i-th focused data type after excluding the g-th non-focused data type from the current monitoring period. Indicates the number of focused data types.

[0058] The greater the difference in browsing concentration changes, the more serious the impact on the overall browsing concentration of all data types before and after the non-focused data type is eliminated, the more important the non-focused data type is, and relatively speaking, the higher the degree of attention diversion of the non-focused data type.

[0059] Step S2-6: Based on the interspersed search of each non-focused data type during the focused data type search in the current monitoring time period, combined with the difference in browsing focus changes, the degree of attention diversion of each non-focused data type to the focused data type is obtained.

[0060] During the current monitoring period, searches for non-focused data types may be interspersed during searches for focused data types. The more serious the interspersed searches for non-focused data types are, the more serious the attention shift from non-focused data types to focused data types is. Combined with the differences in changes in browsing concentration, the degree of attention shift from non-focused data types to focused data types is obtained.

[0061] In an exemplary embodiment, Figure 5As shown, a specific process of obtaining the degree of attention shift is given as follows: Step S2-6-1: Obtain the proportion of searches for each non-focused data type during the focused data type search process in the current monitoring time period.

[0062] The g-th non-focused data type represents any non-focused data type, and the i-th focused data type represents any focused data type. In the current monitoring time period, during the search process for the i-th focused data type, the number of searches for the g-th non-focused data type is obtained, thereby obtaining the number of searches for the g-th non-focused data type in each search process for the focused data type. For all focused data types, the sum of the number of searches for the g-th non-focused data type is calculated. Then, for all focused data types, the sum of the number of searches for all non-focused data types is calculated, and finally, the ratio of the sum of the number of searches for the g-th non-focused data type to the sum of the number of searches is calculated as the proportion of searches for the g-th non-focused data type.

[0063] Therefore, a greater proportion of searches for the gth non-focused data type indicates a higher degree of interspersed searches for the gth non-focused data type during searches for the focused data type, and a higher degree of attention shift for the gth non-focused data type. The proportion of searches is directly proportional to the degree of attention shift.

[0064] Step S2-6-2: Obtain the search duration ratio of each non-focused data type during the focused data type search in the current monitoring time period.

[0065] In the current monitoring time period, during the search process for the i-th focused data type, the search duration for the g-th non-focused data type is obtained, thereby obtaining the search duration for the g-th non-focused data type in each focused data type search process. For all focused data types, the sum of the search durations for the g-th non-focused data type is calculated. Then, for all focused data types, the sum of the search durations for all non-focused data types is calculated. Finally, the ratio of the sum of the search durations for the g-th non-focused data type to the sum of the search durations is calculated as the search duration ratio for the g-th non-focused data type.

[0066] Therefore, a greater proportion of search time for the gth non-focused data type indicates a higher degree of interspersed searches for the gth non-focused data type during searches for the focused data type, and a higher degree of attention shift for the gth non-focused data type. The proportion of search time is directly proportional to the degree of attention shift.

[0067] Step S2-6-3: According to the difference in the proportion of search times, the proportion of search time and the change in browsing concentration, the degree of attention transfer from each non-focused data type to the focused data type is obtained.

[0068] Based on the difference in search frequency, search duration, and browsing focus for the g-th non-focused data type, we determined the degree of attention shift from the g-th non-focused data type to the focused data type. The degree of attention shift is proportional to the difference in search frequency, search duration, and browsing focus.

[0069] In an exemplary embodiment, a specific quantitative method for the degree of attention diversion of the g-th non-focused data type is given as follows: the product of the proportion of search times, the proportion of search time and the difference in browsing concentration change of the g-th non-focused data type is calculated, and the result obtained is the degree of attention diversion of the g-th non-focused data type to the focused data type.

[0070] Step S2-7: Based on the degree of attention transfer and browsing concentration of each non-focused data type, the degree of new consideration of the portrait of each non-focused data type is obtained.

[0071] The higher the degree of attention shift and importance of a non-focused data type, the more it warrants consideration for adding it to a profile, and the higher its profile consideration level. The higher the browsing focus and importance of a non-focused data type, the more it warrants consideration for adding it to a profile, and the higher its profile consideration level. Therefore, the profile consideration level is proportional to both the degree of attention shift and the browsing focus.

[0072] In an exemplary embodiment, a specific method for quantifying the degree of newly added consideration for the portrait of the g-th non-focused data type is provided as follows: the product of the degree of attention shift of the g-th non-focused data type and the browsing focus of the g-th non-focused data type is calculated, and then the product is normalized. The normalized result is the degree of newly added consideration for the portrait of the g-th non-focused data type. The normalization method here is: obtaining the maximum and minimum values ​​of the product of the degree of attention shift of various non-focused data types and their browsing focus, and then normalizing the product of the degree of attention shift of the g-th non-focused data type and the browsing focus of the g-th non-focused data type using the maximum and minimum value normalization method.

[0073] Step S2-8: The non-focused data type corresponding to the new portrait consideration level that is greater than the preset new portrait consideration level threshold is used as the target data type.

[0074] This embodiment predefines a threshold for the degree of newly considered portraits. This threshold is used to determine whether the degree of newly considered portraits of various non-focused data types is high, thereby completing the screening of non-focused data types. It should be understood that the value range of this threshold is 0-1. Provided that the above determination requirements are met, the specific value of this threshold can be flexibly set by the implementer, such as 0.7.

[0075] Compare the new portrait consideration levels of various non-focused data types with the preset portrait new consideration level threshold, obtain the non-focused data type corresponding to the portrait new consideration level that is greater than the preset portrait new consideration level threshold, and use the non-focused data type obtained here as the target data type, thereby completing the addition and update of the target data type.

[0076] Therefore, in addition to directly filtering the target data type through the above browsing focus, for data types whose browsing focus does not meet the conditions, the above process is used to complete the addition of target data types to improve the accuracy of obtaining the target data type and avoid missing data types.

[0077] Step S3: Based on the change trend of the browsing concentration of the target data type in multiple consecutive monitoring time periods, the increase degree of the browsing concentration is obtained.

[0078] Because users' attention to different things often fluctuates, user profiles should also evolve accordingly. When a user's browsing focus on a certain data type increases over multiple consecutive monitoring periods, the user's preference for that data type should be increased, meaning the importance of the profile should be increased. Conversely, the importance of the profile should be reduced. Furthermore, the closer the user's last search for this data type is to the current time, and the greater the corresponding browsing focus, the more emphasis should be placed on the user's attention to this data type in the profile.

[0079] Step S2 is used to obtain the browsing concentration of various target data types in each monitoring time period. The following is an explanation using any target data type as an example. Then, the browsing concentration of the target data type in multiple consecutive monitoring time periods is obtained, and then the browsing concentration of the target data type in each monitoring time period is arranged in chronological order to obtain a browsing concentration sequence of the target data type. According to the changing trend of the browsing concentration in the browsing concentration sequence of the target data type, the degree of increase in the browsing concentration of the target data type is obtained.

[0080] In an exemplary embodiment, Figure 6 As shown, a specific process of obtaining the degree of increase in browsing concentration is given as follows: Step S3-1: Obtain the proportion of the number of monitoring time periods in the first monitoring time period.

[0081] This embodiment sets a judgment logic: for any monitoring time period, if the browsing concentration of the target data type in the monitoring time period is greater than the browsing concentration of the target data type in the previous monitoring time period, it indicates that the browsing concentration of the target data type in the monitoring time period is increasing, and the monitoring time period is defined as the first monitoring time period. In other words, the first monitoring time period refers to a monitoring time period in which the browsing concentration of the target data type is greater than that in the previous monitoring time period.

[0082] Traverse all monitoring time periods and screen each monitoring time period to determine whether each monitoring time period is the first monitoring time period. Then obtain the number of first monitoring time periods. Calculate the ratio of the number of first monitoring time periods to the total number of monitoring time periods as the proportion of the number of monitoring time periods in the first monitoring time period. The greater the proportion of the number of monitoring time periods in the first monitoring time period, the greater the increase in the browsing concentration of the target data type. The increase in browsing concentration is proportional to the proportion of the number of monitoring time periods.

[0083] It should be understood that since there is no monitoring time period before the first monitoring time period in the time sequence, the first monitoring time period does not participate in the comparison of the browsing concentration of the target data type with the previous adjacent monitoring time period.

[0084] Step S3-2: Obtain an average value of the difference between the browsing concentration of the target data type in all first monitoring time periods and the previous adjacent monitoring time period.

[0085] For any first monitoring time period, calculate the difference in browsing concentration of the target data type between the first monitoring time period and the previous monitoring time period adjacent to it. Then calculate the average value of the difference in browsing concentration of the target data type between each first monitoring time period and the previous monitoring time period adjacent to it. The larger the average value of the difference, the more obvious the increase in browsing concentration of the target data type, that is, the greater the increase in browsing concentration of the target data type. The degree of increase in browsing concentration is proportional to the average value of the difference.

[0086] Step S3-3: Obtain the degree of increase in browsing concentration based on the proportion of the number of monitoring time periods and the average value of the difference.

[0087] According to the proportion of the number of monitoring time periods of the target data type and the average value of the difference corresponding to the target data type, the degree of increase in browsing concentration of the target data type is obtained. In an exemplary embodiment, a quantitative method for the increase in browsing concentration is given as follows: the product of the proportion of the number of monitoring time periods of the target data type and the average value of the difference corresponding to the target data type is calculated, and then normalized, and the result obtained is the degree of increase in browsing concentration of the target data type. The normalization method here can be a sigmoid function. The greater the increase in browsing concentration of the target data type, the higher the importance of the portrait of the target data type. The importance of the portrait is proportional to the increase in browsing concentration.

[0088] Step S4: According to the increase in the browsing concentration, combined with the browsing concentration of the target data type and the most recent search time interval of the target data type, the importance of the portrait of the target data type is obtained.

[0089] In this embodiment, the target historical monitoring time period is set to the previous monitoring time period adjacent to the current monitoring time period. The search end time of the target data type in the target historical monitoring time period is obtained, thereby obtaining the time interval between the search end time of the target data type in the target historical monitoring time period and the end time of the current monitoring time period, which is the most recent search time interval of the target data type. The shorter the most recent search time interval of the target data type, the closer the most recent search distance of the target data type is to the current one, and the higher the importance of the portrait of the target data type. Then, the importance of the portrait of the target data type is inversely proportional to the most recent search time interval of the target data type.

[0090] Moreover, the higher the browsing concentration of the target data type in the current monitoring period, the more important the portrait of the target data type is. The importance of the portrait of the target data type is proportional to the browsing concentration in the current monitoring period.

[0091] In an exemplary embodiment, a quantitative method for the portrait importance of the target data type is given as follows: the recent search time interval of the target data type is negatively normalized, and then the product of the increase in browsing concentration of the target data type, the browsing concentration of the target data type in the current monitoring time period, and the recent search time interval of the target data type after negative correlation normalization is calculated, and the obtained product is used as the portrait importance feature of the target data type. Finally, the portrait importance feature of the target data type is normalized, and the normalized result is the portrait importance of the target data type. The normalization method here can be: obtain the maximum and minimum values ​​in the portrait importance features of each target data type, and then use the maximum and minimum value normalization method to normalize the portrait importance features of the target data type.

[0092] Using the above process, the importance of the profile for each target data type is obtained. The user ID, the user data for each target data type, and the importance of the profile for each target data type are then mapped and stored in a database. The user data for each target data type can be retrieved from the database using an SQL query statement. The user ID, the user data for each target data type, and the importance of the profile for each target data type are then visualized, as shown in Table 1, which shows an example of the user profile visualization results.

[0093] Table 1 In addition, target data types with a profile importance greater than a preset value, such as 0.5, can be selected as a reference for user preferences when constructing user profiles. At the same time, the KMP (Knuth-Morris-Pratt) matching algorithm is used to match data such as name, ID, and gender from the user's personal information fields, and the user's preferred data types are stored in association with their user information.

[0094] 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.

[0095] 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 method for generating Internet user portraits based on big data, characterized by: include: Obtain the number of searches and search depths for the target data type obtained by browsing web pages during the user monitoring period, and obtain the progressive search performance of the target data type; Obtaining browsing focus on the target data type based on the search progression performance and in combination with the interspersed search conditions of other data types during the search for the target data type; Obtaining a degree of increase in browsing concentration based on a change trend of browsing concentration of the target data type within multiple consecutive monitoring time periods; According to the increase degree of the browsing concentration, combined with the browsing concentration of the target data type and the most recent search time interval of the target data type, the portrait importance of the target data type is obtained.

2. The method for generating an Internet user portrait based on big data according to claim 1, wherein: The process of obtaining the search progressive representation includes: Obtain the search frequency ratio of the target data type; Obtaining a search depth difference between the search depth and a preset search depth; According to the proportion of the number of times and the difference in the search depth, a search progressive performance of the target data type is obtained, and the search progressive performance is proportional to the proportion of the number of times and inversely proportional to the difference in the search depth.

3. The method for generating an Internet user portrait based on big data according to claim 1, wherein: The process of obtaining the interspersed search situation includes: Obtain the number of data types of other data types interspersed between any two adjacent search processes for the target data type, as well as the interspersed search duration; Obtaining a concentration coefficient corresponding to any two search processes according to the number of data types and the interspersed search duration; the concentration coefficient is inversely proportional to the number of types and the interspersed search duration; The concentration coefficients corresponding to any two search processes are combined to obtain the comprehensive concentration coefficient.

4. The method for generating an Internet user portrait based on big data according to claim 3, wherein: The process of obtaining the browsing concentration of the target data type includes: The browsing concentration is obtained by integrating the comprehensive concentration coefficient and the search progressive performance.

5. The method for generating an Internet user portrait based on big data according to claim 1, wherein: The process of obtaining the degree of increase in browsing concentration includes: Obtaining a ratio of the number of monitoring time periods in a first monitoring time period, where the first monitoring time period refers to a monitoring time period having a greater browsing concentration on the target data type than an adjacent previous monitoring time period; Obtaining an average value of the difference between the browsing concentration of the target data type in all first monitoring time periods and the previous adjacent monitoring time period; The degree of increase in browsing concentration is obtained according to the proportion of the number of monitoring time periods and the average value of the difference. The degree of increase in browsing concentration is proportional to the proportion of the number of monitoring time periods and the average value of the difference.

6. The method for generating an Internet user portrait based on big data according to claim 1, wherein: The monitoring time period includes a plurality of historical monitoring time periods and a current monitoring time period arranged in chronological order; the most recent search time interval of the target data type is the time interval between the end time of the search for the target data type in the target historical monitoring time period and the end time of the current monitoring time period; the target historical monitoring time period is the previous monitoring time period adjacent to the current monitoring time period; The importance of the portrait is directly proportional to the increase in the browsing concentration and the browsing concentration of the target data type in the current monitoring time period, and is inversely proportional to the most recent search time interval.

7. The method for generating an Internet user portrait based on big data according to claim 1, wherein: The monitoring time period includes a plurality of historical monitoring time periods and a current monitoring time period arranged in chronological order; The process of obtaining the target data type includes: Based on browsing focus, the initial data types are divided into focused data types and non-focused data types; The initial data type includes the target data type, and the target data type includes the focused data type; Obtain the difference in browsing concentration of the focused data type before and after excluding each non-focused data type from the current monitoring time period; According to the interspersed search situation of each non-focused data type during the focused data type search process in the current monitoring time period, combined with the difference in the browsing focus change, the degree of attention diversion of each non-focused data type to the focused data type is obtained; Obtaining a new consideration level for a portrait of each non-focused data type based on the attention shift level and browsing concentration level of each non-focused data type; The non-focused data type corresponding to the new portrait consideration level that is greater than the preset new portrait consideration level threshold is used as the target data type.

8. The method for generating an Internet user portrait based on big data according to claim 7, wherein: The process of acquiring the degree of attention shifting includes: Get the percentage of searches for each non-focused data type during the current monitoring period. Get the search duration ratio of each non-focused data type during the focused data type search in the current monitoring time period; Based on the proportion of search times, the proportion of search duration and the difference in browsing concentration changes, the degree of attention diversion of each non-focused data type to the focused data type is obtained; the degree of attention diversion is proportional to the proportion of search times, the proportion of search duration and the difference in browsing concentration changes.

9. The method for generating an Internet user portrait based on big data according to claim 7, wherein: The initial data types are divided into focused data types and non-focused data types based on the browsing focus, including: Comparing the browsing concentration of each initial data type during the current monitoring period with a preset browsing concentration threshold; An initial data type corresponding to a browsing concentration greater than the preset browsing concentration threshold is determined as the concentration data type.

10. The method for generating Internet user portraits based on big data according to claim 1, wherein: After obtaining the portrait importance of the target data type, the big data-based Internet user portrait generation method further includes: visualizing the user ID, user data of the target data type, and the portrait importance.

Images