An intelligent analysis system based on multidimensional data portrait
Through the multi-dimensional data portrait intelligent analysis system, the update cycle and collection dimension of user portrait labels are dynamically adjusted, which solves the problem of inaccurate user portrait label evaluation, achieves more accurate and efficient label evaluation and maintenance, and improves the accuracy of user portrait identification Spend.
Patent Information
- Application Number
- CN202410738550.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-07
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-06-07
AI Technical Summary
The prior art is difficult to capture inaccuracy and insufficient maintenance strategies of user portrait labels caused by changes in user behavior and market environment in real time, resulting in inaccurate or biased label evaluation results.
Design an intelligent analysis system based on multi-dimensional data portraits, including basic portrait database, portrait update driver unit, portrait analysis unit and portrait data dynamic analysis unit. By monitoring the change frequency and relative change rate of user portrait data, adjust the update cycle and collection Dimensions, dynamically adjust label evaluation and maintenance strategies.
It improves the accuracy and timeliness of label evaluation, reduces computing resource consumption, improves the accuracy of user portrait discrimination, adapts to data changes and business needs, and avoids resource waste and data redundancy.
Smart Images

Figure CN118733637B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of big data and artificial intelligence technology, and in particular to an intelligent analysis system based on multidimensional data portraits. Background Art
[0002] User portraits are a set of multi-dimensional user features that are formed by abstracting and summarizing users based on a series of collected user data and information. These features are usually used to describe user attributes, behaviors, preferences, needs, etc. Tags are the basic components of user portraits, which are specific and quantifiable sets of user features in these sets. Tags can divide users into different groups or types, which is of great significance for accurately positioning users, optimizing user experience, guiding product decisions, improving operational efficiency, and supporting data analysis. Therefore, when building user portraits, we need to pay full attention to the role of tags and reasonably set and update tags according to actual business needs.
[0003] Chinese patent application publication number: CN115062789A, discloses a user portrait label evaluation method, including: constructing an intelligent creative component according to the label to be evaluated; using an automated foundation engine to distribute the intelligent creative component to obtain an experimental group and a control group; performing data mining on the click behavior of the target group to calculate the actual number of click users A; quantifying the indicators of the experimental group and the control group to obtain the accuracy and importance of the label and perform a portrait label evaluation to infer whether the label is important to the corresponding business, that is, whether it is necessary to continuously maintain the label.
[0004] It can be seen that the above technical solution not only evaluates based on click-through rate and accuracy, but also mines and analyzes the data of the experimental group and the control group to evaluate the accuracy and importance of the label, making the label evaluation more comprehensive and accurate. However, there are still the following problems: for newly introduced labels or user groups, there may be inaccurate or biased evaluation results due to the lack of sufficient data to train the model or verify the effect of the label in the early stage. At the same time, as user behavior and market environment change, the accuracy and importance of the label may change. However, the above technical solution may not be able to capture these changes in real time and dynamically adjust the label evaluation and maintenance strategy. Summary of the invention
[0005] To this end, the present invention provides an intelligent analysis system based on multidimensional data portraits to overcome the problem of inaccurate user portrait labels in the prior art.
[0006] To achieve the above objectives, on the one hand, the present invention provides an intelligent analysis system based on multidimensional data portraits, comprising:
[0007] The basic portrait database stores portrait data of several dimensions of users and a portrait label dataset of each user;
[0008] A portrait update driving unit, which is connected to the basic portrait database, and is used to determine the update period of the portrait tag of a single user according to the portrait data of each dimension of the single user, and to determine the single trigger update timing of the portrait tag of the single user according to the relative change rate of the portrait data of any dimension in the portrait data of each dimension of the single user and the update period;
[0009] A portrait analysis unit, which is connected to the basic portrait database and the portrait update driving unit respectively, and is used to adjust the update cycle according to the triggering frequency of a single user portrait tag;
[0010] A portrait data dynamic analysis unit, which is connected to the basic portrait database and is used to calculate the data fluctuation value of the portrait data of each dimension of a single user, determine the contribution degree of the portrait data to the portrait label of the single user according to the data fluctuation value, and determine whether to adjust the collection dimension of the portrait data of the single user in the basic portrait database according to the contribution degree;
[0011] A portrait execution unit is respectively connected to the basic portrait database, the portrait update driving unit and the portrait data dynamic analysis unit, and is used to execute user portraits according to the update cycle, the single trigger update timing and the collection dimension to generate the portrait label data set corresponding to each user.
[0012] Furthermore, the portrait update driving unit determines the update period of the single user portrait label according to the change frequency of the portrait data of each dimension of the single user.
[0013] The change frequency is inversely correlated with the period length of the update period.
[0014] Furthermore, the portrait update driving unit determines a single trigger update timing of a single user portrait tag according to a relative change rate of portrait data of any dimension in the portrait data of each dimension of a single user and the update cycle, including:
[0015] During the update cycle, if the relative change rate of the portrait data in any dimension is greater than the preset relative change rate, it is determined that the portrait data of the single user meets the data conditions of a single trigger update opportunity.
[0016] Furthermore, the portrait analysis unit adjusts the update period according to the triggering frequency of each portrait tag of a single user, including:
[0017] If the trigger frequencies of the portrait tags of a single user are all less than the first set threshold, it is determined to increase the period length of the update period;
[0018] If the trigger frequency of at least one portrait tag among the portrait tags of a single user is greater than or equal to the first set threshold, it is determined to reduce the period length of the update period.
[0019] Furthermore, the single trigger update timing is adjusted, including:
[0020] The portrait update driving unit determines whether the portrait data of a single user meets the time condition of a single trigger update opportunity according to the interval between the judgment result of the data condition and the last portrait tag update, including:
[0021] If the portrait data of a single user is currently judged to meet the data conditions of a single trigger update timing and the interval between the judgment result of the data conditions and the last portrait tag update is longer than the preset waiting interval, the portrait data of a single user is judged to meet the time conditions of a single trigger update timing.
[0022] Furthermore, the portrait data dynamic analysis unit calculates a data fluctuation value of the portrait data of a single dimension of a single user according to a standard deviation of the portrait data of a single dimension of a single user within a set time period.
[0023] Furthermore, the portrait data dynamic analysis unit adjusts the baseline weight of the portrait data of the corresponding dimension according to the calculated data fluctuation value and determines the contribution of each dimension to the portrait label of a single user according to the adjusted weight of each dimension.
[0024] Furthermore, the portrait data dynamic analysis unit is provided with a weight adjustment rule, including:
[0025] If the data fluctuation value is less than the first fluctuation threshold, the contribution degree is the sum of the base weight and the weight increment;
[0026] If the data fluctuation value is greater than the first fluctuation threshold and less than the second fluctuation threshold, the contribution degree is the reference weight;
[0027] If the data fluctuation value is greater than the second fluctuation threshold, the contribution degree is the difference between the reference weight and the weight increment.
[0028] Furthermore, the portrait data dynamic analysis unit determines whether to adjust the collection dimension of the portrait data of a single user in the basic portrait database according to the contribution degree, including:
[0029] If the contribution degree is greater than the contribution degree threshold, retain the collection dimension and continue to collect the portrait data of the collection dimension;
[0030] If the contribution degree is less than or equal to the contribution degree threshold, remove the collection dimension and stop collecting the portrait data of the collection dimension.
[0031] Furthermore, the portrait execution unit executes user portrait according to the update cycle, the single trigger update timing, and the collection dimension to generate the portrait label data set corresponding to each user, including:
[0032] The portrait execution unit re-collects the portrait data under the corresponding dimension according to the collection dimension based on the update cycle or the single trigger update timing and stores it in the basic portrait database;
[0033] The basic portrait database generates the updated portrait label data set corresponding to the user based on the update instruction of the portrait execution unit and the portrait data.
[0034] Compared with the prior art, the beneficial effect of the present invention lies in that the update period and single trigger update timing of the portrait label of a single user are determined and adjusted through the portrait update drive unit and the portrait analysis unit. The update period is determined based on the portrait data of each dimension of the single user and adjusted based on the trigger frequency of the portrait label of the single user. The single trigger update timing must meet the data conditions and time conditions to trigger the portrait label update. The portrait data dynamic analysis unit determines the contribution of the portrait data of each dimension of the single user to the portrait label of the single user according to the data fluctuation value of the portrait data of each dimension of the single user, and adjusts the collection dimension of the portrait data; the portrait label data set corresponding to each user is regenerated through the portrait execution unit. The present invention makes the label evaluation more comprehensive and accurate, and can improve the accuracy of judging user portraits.
[0035] Furthermore, the portrait update driving unit of the present invention flexibly determines the update cycle of a single user portrait label by monitoring the change frequency of the portrait data of each dimension of a single user within a specified time period, thereby avoiding the waste of resources caused by frequent updates of portrait data, greatly reducing the consumption of computing resources, improving the operating efficiency of the overall system, making label evaluation more comprehensive and accurate, and thus improving the accuracy of judging user portraits.
[0036] Furthermore, the portrait update driving unit of the present invention triggers the immediate update of the label or the regular update according to the preset update cycle by monitoring the relative change rate of the portrait data of any dimension in the various dimensions of the portrait data of a single user within the update cycle. This flexible triggering mechanism enables the system to respond quickly to significant changes in the user's portrait data, avoiding inaccurate labels due to delays in portrait data collection, ensuring timeliness while making label evaluation more comprehensive and accurate, thereby improving the accuracy of identifying user portraits.
[0037] Furthermore, the portrait analysis unit of the present invention adjusts the update cycle through the trigger frequency of each portrait tag of a single user, so that changes in the user's portrait data can be responded to more quickly and accurately, which significantly improves data processing efficiency, reduces unnecessary data processing burden, makes label evaluation more comprehensive and accurate, and thus improves the accuracy of identifying user portraits.
[0038] Furthermore, the portrait update driving unit of the present invention determines whether the portrait data of a single user meets the time condition of a single trigger update opportunity based on the interval length from the judgment result of the data condition to the last portrait label update, and triggers the single trigger update opportunity and updates the portrait data when the time condition and the data condition are met, thereby ensuring that the portrait label is not too outdated and at the same time reducing computer pressure, making label evaluation more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0039] Furthermore, the portrait update driving unit of the present invention calculates the data fluctuation value of the portrait data of a single dimension of a single user according to the standard deviation of the portrait data of a single dimension of a single user within a set time period, and judges the stability of the portrait data of the user according to the data fluctuation value. The more unstable the portrait data is, the higher the data fluctuation value is, and vice versa. By judging the stability of the user's portrait data, the label evaluation is made more comprehensive and accurate, thereby improving the accuracy of judging the user portrait.
[0040] Furthermore, the dynamic analysis unit of the portrait data of the present invention adjusts the reference weight of the portrait data of the corresponding dimension according to the calculated data fluctuation value and determines the contribution of each dimension to the portrait label of a single user according to the adjusted weight of each dimension. This means that when assigning or updating labels to users, not only the total amount or frequency of the data is considered, but also the change trend and stability of the data. This makes the label evaluation more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0041] Furthermore, the dynamic analysis of portrait data in the present invention is provided with weight adjustment rules, which dynamically adjust the weights of portrait data of each dimension based on data fluctuation values and weight adjustment rules. This dynamic adjustment method can better adapt to changes in data and changes in business needs, making label evaluation more comprehensive and accurate, thereby improving the accuracy of identifying user portraits.
[0042] Furthermore, the portrait data dynamic analysis unit of the present invention adjusts the collection dimension of the portrait data of a single user based on the degree of contribution. By dynamically adjusting the collection dimension of the user portrait data, the accuracy and efficiency of portrait data collection are improved, the quality and availability of data in the basic portrait database are improved, the waste of resources is avoided, the optimal allocation of resources is achieved, and the label evaluation is more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0043] Furthermore, the portrait execution unit of the present invention re-collects the portrait data under the corresponding dimension according to the update cycle or the single trigger update timing, and the basic portrait database generates an updated portrait label data set of the corresponding user based on the portrait data. The real-time and accuracy of the user portrait data is ensured, and the portrait data under the collection dimension is collected in a targeted manner, which avoids unnecessary data redundancy, saves computing resources, makes the label evaluation more comprehensive and accurate, and thus improves the accuracy of judging user portraits. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a connection block diagram of the intelligent analysis system based on multi-dimensional data portrait of the present invention;
[0045] Figure 2 A schematic diagram of determining a single triggering opportunity of a single user portrait tag in the present invention;
[0046] Figure 3 A schematic diagram of the update cycle of a single user portrait tag after a single trigger update opportunity is triggered in the present invention;
[0047] Figure 4 A schematic diagram of a single user portrait tag set in the updated basic portrait database of the present invention. DETAILED DESCRIPTION
[0048] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0049] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.
[0050] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.
[0051] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0052] See also Figure 1 As shown, it is a connection block diagram of an intelligent analysis system based on multidimensional data portraits of the present invention. The present invention provides an intelligent analysis system based on multidimensional data portraits, including:
[0053] The basic portrait database stores portrait data of several dimensions of users and a portrait label dataset of each user;
[0054] A portrait update driving unit, which is connected to the basic portrait database, and is used to determine the update period of the portrait tag of a single user according to the portrait data of each dimension of the single user, and to determine the single trigger update timing of the portrait tag of the single user according to the relative change rate of the portrait data of any dimension in the portrait data of each dimension of the single user and the update period;
[0055] A portrait analysis unit, which is connected to the basic portrait database and the portrait update driving unit respectively, and is used to adjust the update cycle according to the triggering frequency of a single user portrait tag;
[0056] A portrait data dynamic analysis unit, which is connected to the basic portrait database and is used to calculate the data fluctuation value of the portrait data of each dimension of a single user, determine the contribution degree of the portrait data to the portrait label of the single user according to the data fluctuation value, and determine whether to adjust the collection dimension of the portrait data of the single user in the basic portrait database according to the contribution degree;
[0057] A portrait execution unit is respectively connected to the basic portrait database, the portrait update driving unit and the portrait data dynamic analysis unit, and is used to execute user portraits according to the update cycle, the single trigger update timing and the collection dimension to generate the portrait label data set corresponding to each user.
[0058] It is understandable that the intelligent analysis system stores the portrait data of several dimensions of the user and the portrait label data set of each user in the basic portrait database. The portrait update driving unit determines the update cycle and single trigger update timing of the portrait label of a single user based on the portrait data of each dimension of the single user. The single trigger update timing must meet the data conditions and time conditions to trigger the portrait label update. The portrait analysis unit adjusts the update cycle based on the trigger frequency of the portrait label of a single user; then the portrait data dynamic analysis unit analyzes the data fluctuation value of the portrait data of each dimension of a single user, and determines the adjustment strategy of the portrait data collection dimension. The portrait execution unit and the basic portrait database execute the construction and update of the user portrait according to the update cycle, trigger timing and collection dimension, and generate a new portrait label data set.
[0059] The present invention determines and adjusts the update cycle and single trigger update timing of the portrait label of a single user through a portrait update driving unit and a portrait analysis unit. The update cycle is determined based on the portrait data of each dimension of a single user and adjusted based on the trigger frequency of the portrait label of a single user. The single trigger update timing must meet data conditions and time conditions to trigger the portrait label update. The portrait data dynamic analysis unit determines the contribution of the portrait data of each dimension of a single user to the portrait label of a single user according to the data fluctuation value of the portrait data of each dimension of a single user, and adjusts the collection dimension of the portrait data; the portrait label data set corresponding to each user is regenerated through the portrait execution unit. The present invention makes the label evaluation more comprehensive and accurate, and can improve the accuracy of judging user portraits.
[0060] Specifically, the portrait update driving unit determines the update period of the single user portrait label according to the change frequency of the portrait data of each dimension of the single user.
[0061] The change frequency is inversely correlated with the period length of the update period.
[0062] It can be understood that the change frequency is the number of times the portrait data of each dimension of a single user changes within a unit time length. For example, if the number of times the portrait data of each dimension of a single user changes within a unit time length is 10, then the change frequency is 10.
[0063] The intelligent analysis system will continuously monitor the frequency of changes in the portrait data within a set time period and dynamically determine the update cycle:
[0064] When the portrait data change frequency is at a high level and higher than the first change frequency, the update cycle is the first update cycle. The first update cycle is shorter and is used to quickly respond to changes in the user's portrait data to ensure the real-time nature of the user portrait.
[0065] When the portrait data change frequency is at a medium level and is higher than the second change frequency but lower than the first change frequency, the update period is the second update period, which is slightly longer than the first update period to reduce unnecessary update frequency while maintaining a certain accuracy.
[0066] When the frequency of portrait data change is zero, that is, the user portrait data has not changed, the intelligent analysis system will use the longest update cycle, that is, the third update cycle, or choose to wait until the portrait data changes before updating. The third update cycle is longer, which can save system resources because there is no need to perform frequent updates when the user portrait data is stable.
[0067] In implementation, the first change frequency ranges from 10 to 15 times / cycle, preferably, the first change frequency is 10 times / cycle, the second change frequency ranges from 5 to 10 times / cycle, preferably, the second change frequency is 5 times / cycle, preferably, the set time period is one week, the first update cycle is one day, the second update cycle is one week, and the third update cycle is three weeks or one month.
[0068] The portrait update driving unit of the present invention flexibly determines the update cycle of a single user portrait label by monitoring the change frequency of the portrait data of each dimension of a single user within a specified time period, thereby avoiding the waste of resources caused by frequent updates of portrait data, greatly reducing the consumption of computing resources, improving the operating efficiency of the overall system, making label evaluation more comprehensive and accurate, and thus improving the accuracy of judging user portraits.
[0069] Specifically, the portrait update driving unit determines a single trigger update timing of a single user portrait tag according to a relative change rate of any dimension of the portrait data of each dimension of a single user and the update cycle, including:
[0070] During the update cycle, if the relative change rate of the portrait data in any dimension is greater than the preset relative change rate, it is determined that the portrait data of the single user meets the data conditions of a single trigger update opportunity.
[0071] It can be understood that the relative change rate is the relative change rate of the portrait data of any dimension within a specific time interval within the update cycle (such as from the first time point to the second time point). The relative change rate indicator can more accurately reflect the changes in the portrait data of any dimension of the user. During the update cycle, if the relative change rate of the portrait data of any dimension at the first time point and the second time point is greater than the preset relative change rate, it is determined that the portrait data of a single user meets the data conditions of a single trigger update opportunity. If the relative change rate of the portrait data of any dimension at the first time point and the second time point is less than or equal to the preset relative change rate, it means that the user's portrait data is relatively stable and there is no need to frequently trigger updates. Each user portrait label can be updated according to the update cycle, thereby improving system efficiency and reducing unnecessary waste of resources.
[0072] In a specific embodiment, the relative change rate of the numerical dimension data The calculation formula is as follows:
[0073] (1)
[0074] in, For the second time point image data, The image data is the first point in time.
[0075] In another specific embodiment, the relative change rate of the descriptive dimensional data (other image data except the numerical dimensional data) The calculation formula is as follows:
[0076] (2)
[0077] The descriptive dimension data of a certain dimension of the single user is set to A1 at the first time point and to A2 at the second time point. is the number of users whose descriptive dimension data is A1 at the first time point among all users, The number of users whose descriptive dimension data is A1 at the first time point and whose descriptive dimension data changes to A2 at the second time point. For example, the music style and rhythm dimension data of a single user is cheerful at the first time node and soothing at the second time node, and there are 100 users whose music style and rhythm are cheerful at the first time node, and 5 of the 100 users whose music style and rhythm change to soothing at the second time node, then is 100, is 5.
[0078] In implementation, the first time point and the second time point are determined based on the update cycle. If the update cycle is 1 day, preferably, the first time point is 9:00 and the second time point is 13:00; if the update cycle is 1 week, preferably, the first time point is Thursday and the second time point is Friday; if the update cycle is January, preferably, the first time point is the 11th and the second time point is the 17th; the preset relative change rate ranges from 30% to 50%, preferably, the preset relative change rate is 30%. The value ranges and preferred values of the first time point, the second time point, and the preset relative change rate should be determined according to actual conditions and will not be repeated here.
[0079] The portrait update driving unit of the present invention triggers the immediate update of the label or the regular update according to the preset update cycle by monitoring the relative change rate of the portrait data of any dimension in the portrait data of each dimension of a single user within the update cycle. This flexible trigger mechanism enables the system to quickly respond to significant changes in the user's portrait data, avoiding inaccurate labels due to delays in portrait data collection, ensuring timeliness while making label evaluation more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0080] Specifically, the portrait analysis unit adjusts the update period according to the triggering frequency of each portrait tag of a single user, including:
[0081] If the trigger frequencies of the portrait tags of a single user are all less than the first set threshold, it is determined to increase the period length of the update period;
[0082] If the trigger frequency of at least one portrait tag among the portrait tags of a single user is greater than or equal to the first set threshold, it is determined to reduce the period length of the update period.
[0083] It can be understood that the trigger frequency is the number of times each portrait tag of a single user is triggered to update due to changes in the portrait data of several dimensions of the user. If the trigger frequency of each portrait tag of a single user is less than the first set threshold, it means that the portrait data of each dimension of the user under the portrait tag does not change frequently. At this time, in order to reduce the computing cost, the cycle length of the update cycle is increased; if the trigger frequency of at least one portrait tag among the portrait tags of a single user is greater than or equal to the first set threshold, it means that the portrait data of each dimension of the user under the portrait tag changes frequently. At this time, in order to ensure the accuracy of the user portrait, the cycle length of the update cycle is reduced.
[0084] In implementation, the first set threshold value ranges from 3 to 5 times, preferably, the first set threshold value is 4, the update cycle increase amount needs to be determined according to the update cycle length, preferably, the update cycle increase amount and the update cycle decrease amount are both half of the update cycle length. The value range and preferred value of the first set threshold value and the update cycle need to be selected according to actual conditions, which will not be repeated here.
[0085] The portrait analysis unit of the present invention adjusts the update cycle through the trigger frequency of each portrait tag of a single user, so that changes in the user's portrait data can be responded to more quickly and accurately, which significantly improves data processing efficiency, reduces unnecessary data processing burden, makes label evaluation more comprehensive and accurate, and thus improves the accuracy of judging user portraits.
[0086] See also Figure 2 and Figure 3 As shown, Figure 2 A schematic diagram of determining a single triggering opportunity of a single user portrait tag in the present invention, Figure 3 The figure is a schematic diagram of the update cycle of a single user portrait tag after a single trigger update opportunity is triggered in the present invention. Specifically, the portrait update driving unit determines whether the portrait data of a single user meets the time condition of the single trigger update opportunity according to the interval between the judgment result of the data condition and the last portrait tag update, including:
[0087] If the portrait data of a single user is currently judged to meet the data conditions of a single trigger update timing and the interval between the judgment result of the data conditions and the last portrait tag update is longer than the preset waiting interval, the portrait data of a single user is judged to meet the time conditions of a single trigger update timing.
[0088] It can be understood that the single trigger update timing must not only meet the data conditions but also the time conditions. The preset waiting interval is set to allow the single trigger update timing to have a waiting period after the last label update cycle, so as to avoid the portrait update drive unit executing the update right after the user portrait label is updated.
[0089] In implementation, the preset waiting interval should be determined according to the update cycle. If the update cycle is 1 day, the preset waiting interval is in the range of 8 to 16 hours, and preferably, the preset waiting interval is 8 hours; if the update cycle is 1 week, the preset waiting interval is in the range of 2 to 4 days, and preferably, the preset waiting interval is 2 days; if the update cycle is 1 month, the preset waiting interval is in the range of 10 to 20 days, and preferably, the preset waiting interval is 10 days; the interval range and preferred value should be determined according to actual conditions, which will not be repeated here.
[0090] The portrait update driving unit of the present invention determines whether the portrait data of a single user meets the time condition of a single trigger update opportunity according to the interval length from the judgment result of the data condition to the last portrait label update, and triggers the single trigger update opportunity and updates the portrait data when the time condition and the data condition are met, thereby ensuring that the portrait label is not too outdated and at the same time reducing the pressure on the computer, making the label evaluation more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0091] Specifically, the portrait data dynamic analysis unit calculates the data fluctuation value of the portrait data of a single dimension of a single user according to the standard deviation of the portrait data of a single dimension of a single user within a set time period.
[0092] It is understandable that the data fluctuation value is obtained by calculating the standard deviation of the portrait data of a single dimension within a set time period. The standard deviation can quantify the degree of discreteness of the data set, that is, the degree of deviation between the data point and the average value. The closer the standard deviation is to 0, the more stable the overall data set. For user portrait data, it means that by using the standard deviation, the degree of fluctuation of the portrait data of each dimension of the user within a set time period can be measured. In implementation, the set time period is the update cycle.
[0093] In a specific embodiment, the data fluctuation value of the numerical dimension data The calculation formula is as follows:
[0094] (3)
[0095] in, is the portrait data in a single dimension. is the mean of the portrait data in a single dimension, It is the total number of portrait data under a single dimension within the set time period.
[0096] In another specific embodiment, the data fluctuation value of the descriptive dimension data (other image data except the numerical dimension data) The calculation formula is as follows:
[0097] (4)
[0098] in, The number of changes in the profile data of a single user in a single dimension within each preset time period. It is the average of all changes in the portrait data of a single user in a single dimension within each preset time period. is the total number of all preset time periods. The preset time period is determined according to the update cycle. For example, if the update cycle is 1 week, the preset time period is 1 day. is 7.
[0099] The portrait update driving unit of the present invention calculates the data fluctuation value of the portrait data of a single dimension of a single user according to the standard deviation of the portrait data of a single dimension of a single user within a set time period, and judges the stability of the portrait data of the user according to the data fluctuation value. The more unstable the portrait data is, the higher the data fluctuation value is, and vice versa. By judging the stability of the user's portrait data, the label evaluation is made more comprehensive and accurate, thereby improving the accuracy of judging the user portrait.
[0100] Specifically, the portrait data dynamic analysis unit adjusts the baseline weight of the portrait data of the corresponding dimension according to the calculated data fluctuation value and determines the contribution of each dimension to the portrait label of a single user according to the adjusted weight of each dimension.
[0101] It is understandable that portrait data of different dimensions have different importance and influence. In order to reflect this difference, a baseline weight is first determined for the portrait data of each dimension, and the baseline weight represents the relative importance of the portrait data of each dimension in the user portrait. At the same time, the baseline weight is adjusted according to the weight adjustment rule, and the weight adjustment rule can be set according to actual needs. The adjusted weight is set as the contribution of the portrait data of each dimension to the portrait label of the single user.
[0102] In implementation, the sum of the benchmark weights of the portrait data of each dimension is 1, and the benchmark weight is determined based on the acquisition dimension. For example, if the acquisition dimension is 5, the benchmark weight is 0.2. The preferred value of the benchmark weight should be adjusted according to actual conditions and will not be described here.
[0103] The dynamic analysis unit of the portrait data of the present invention adjusts the reference weight of the portrait data of the corresponding dimension according to the calculated data fluctuation value and determines the contribution of each dimension to the portrait label of a single user according to the adjusted weight of each dimension. This means that when assigning or updating labels to users, not only the total amount or frequency of the data is considered, but also the change trend and stability of the data are considered. This makes the label evaluation more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0104] Specifically, the image data dynamic analysis unit is provided with weight adjustment rules, including:
[0105] If the data fluctuation value is less than the first fluctuation threshold, the contribution degree is the sum of the base weight and the weight increment;
[0106] If the data fluctuation value is greater than the first fluctuation threshold and less than the second fluctuation threshold, the contribution degree is the reference weight;
[0107] If the data fluctuation value is greater than the second fluctuation threshold, the contribution degree is the difference between the base weight and the weight increment.
[0108] It can be understood that as the data fluctuation value increases, the contribution degree gradually decreases. When the data fluctuation value is less than the first fluctuation threshold, it indicates that the user's portrait data stability on this dimension is high, and therefore the contribution to the portrait label is large, and a weight slightly higher than the benchmark weight is given; if the data fluctuation value is greater than the first fluctuation threshold and less than the second fluctuation threshold, it indicates that the user's portrait data stability on this dimension is moderate, and the contribution to the portrait label is average, and a benchmark weight is given; if the data fluctuation value is greater than the second fluctuation threshold, it indicates that the user's portrait data stability on this dimension is poor, and the contribution to the portrait label is small, and a weight lower than the benchmark weight is given.
[0109] In implementation, the first fluctuation threshold is smaller than the second fluctuation threshold. The selection of the first fluctuation threshold and the second fluctuation threshold is determined based on the obtained data fluctuation value. Preferably, the first fluctuation threshold value should be close to 0, and the second fluctuation threshold value should be greater than 0. For example, the first fluctuation threshold value is one-tenth of the average value of all data fluctuation values of the portrait data of the corresponding dimension within several historical set time periods, and the second fluctuation threshold value is one-fifth of the average value of all data fluctuation values of the portrait data of the corresponding dimension within several historical set time periods. The weight increment for a single adjustment is set to 0.05. The first fluctuation threshold value, the second fluctuation threshold value, and the preferred value of the weight increment need to be selected based on actual conditions, which will not be elaborated here.
[0110] The dynamic analysis of portrait data in the present invention is provided with weight adjustment rules, which dynamically adjust the weights of portrait data of each dimension based on data fluctuation values and weight adjustment rules. This dynamic adjustment method can better adapt to changes in data and changes in business needs, making label evaluation more comprehensive and accurate, thereby improving the accuracy of identifying user portraits.
[0111] Specifically, the portrait data dynamic analysis unit determines whether to adjust the collection dimension of the portrait data of a single user in the basic portrait database according to the contribution degree, including:
[0112] If the contribution degree is greater than the contribution degree threshold, retain the collection dimension and continue to collect the portrait data of the collection dimension;
[0113] If the contribution degree is less than or equal to the contribution degree threshold, remove the collection dimension and stop collecting the portrait data of the collection dimension.
[0114] It can be understood that in order to decide whether to adjust the collection dimension of the portrait data of a single user, the contribution degree threshold is set. If the contribution degree is greater than the contribution degree threshold, it means that the portrait data under this dimension is very important for constructing the user portrait, so it is chosen to retain the collection dimension and continue to collect the portrait data of this collection dimension; if the contribution degree is less than or equal to the contribution degree threshold, it means that the contribution of the portrait data under this dimension to constructing the user portrait is relatively low, so the collection dimension is removed and the collection of portrait data of this collection dimension is stopped.
[0115] In implementation, preferably, the contribution degree threshold is taken as the average of the contribution degrees under the various dimensions. The preferred value of the contribution degree threshold should be selected according to actual conditions and will not be repeated here. Preferably, the contribution degree threshold is set to 0.25 times the benchmark weight.
[0116] The portrait data dynamic analysis unit of the present invention adjusts the collection dimension of the portrait data of a single user based on the degree of contribution. By dynamically adjusting the collection dimension of the user portrait data, the accuracy and efficiency of portrait data collection are improved, the quality and availability of data in the basic portrait database are improved, the waste of resources is avoided, the optimal allocation of resources is achieved, and the label evaluation is more comprehensive and accurate, thereby improving the accuracy of judging user portraits.
[0117] See also Figure 4 The figure is a schematic diagram of a single user portrait label set in the basic portrait database after the present invention is updated. Specifically, the portrait execution unit executes the user portrait according to the update cycle, the single trigger update timing, and the collection dimension to generate the portrait label data set corresponding to each user, including:
[0118] The portrait execution unit re-collects the portrait data under the corresponding dimension according to the collection dimension based on the update cycle or the single trigger update timing and stores it in the basic portrait database;
[0119] The basic portrait database generates the updated portrait label data set corresponding to the user based on the update instruction of the portrait execution unit and the portrait data.
[0120] It can be understood that the portrait execution unit will re-collect the portrait data of each user under the corresponding dimension according to the readjusted collection dimension, and update the portrait data according to the update cycle or the single trigger update timing, so as to generate the portrait label data set corresponding to each user, realize the update and maintenance of the portrait label, and improve the accuracy of the user portrait.
[0121] The portrait execution unit of the present invention re-collects the portrait data under the corresponding dimension according to the collection dimension according to the update cycle or the single trigger update timing, and the basic portrait database generates the updated portrait label data set of the corresponding user based on all the portrait data (the portrait data of each dimension after the dimension adjustment). The real-time and accuracy of the user portrait data is ensured, and the portrait data under the collection dimension is collected in a targeted manner, which avoids unnecessary data redundancy, saves computing resources, makes the label evaluation more comprehensive and accurate, and thus improves the accuracy of judging the user portrait.
[0122] In order to further illustrate the present invention, the intelligent analysis system based on multidimensional data profiling provided by the present invention is described in detail below, but they should not be understood as limiting the scope of protection of the present invention. Example
[0123] Take the following example for user's music preference tags:
[0124] The portrait data of several dimensions in the basic portrait database include: songs listened to, number of songs listened to, time of listening to songs, number of concerts purchased, number of albums purchased, number of song recommendations, user age, number of user comments, user IP address, song region, time period of listening to songs, and music style and rhythm. The user's portrait label data set is shown in the following table:
[0125] Tag Category User 1 User 2 User 3 Music Type Rock ballad Popularity Emotional Type Passion calm Cheerful Scene Type sports rest Work
[0126] User 1's portrait tags are rock, passion, and sports. The portrait data of each dimension under different tags are analyzed, and based on the changes in the portrait data of each dimension, the update cycle of user 1's portrait tag is determined to be 1 month. The first 10 days of the update cycle are the preset waiting interval, the 11th day of the month is the first time point, and the 17th day of the month is the second time point. When the relative change rate of the portrait data of user 1 under the dimension of the song list is greater than the preset relative change rate of 30%, the portrait data of user 1 meets the data conditions of the single trigger update opportunity.
[0127] Determination of collection dimensions: Within the update cycle of 1 month, the number of collection dimensions under the music type label of user 1 is set to 5, and the collection dimensions are respectively, the list of songs listened to, the number of times listened to songs, the time of listening to songs, the number of times concerts purchased, and the number of times albums purchased, and the base weight of each dimension is 0.2. The data fluctuation values under the five dimensions are calculated respectively. The data fluctuation value of the number of songs listened to is less than the preset first fluctuation threshold, so the contribution degree is 0.25, the data fluctuation value of the number of times listened to songs is less than the preset first fluctuation threshold, so the contribution degree is 0.25, the data fluctuation value of the time of listening to songs is greater than the first fluctuation threshold and less than the second fluctuation threshold, so the contribution degree is 0.2, the data fluctuation value of the number of times concerts purchased is greater than the second fluctuation threshold, so the contribution degree is 0.15, and the data fluctuation value of the number of times albums purchased is greater than the second fluctuation threshold, so the contribution degree is 0.15. Set the contribution degree threshold to 0.17, then retain the three collection dimensions of the list of songs listened to, the number of times listened to songs, and the time of listening to songs, and remove the two collection dimensions of the number of times concerts purchased and the number of times albums purchased.
[0128] Regenerate the portrait label data set of user 1: After re-determining the update cycle and collection dimension, determine the new portrait data, and generate a new music category label of folk songs. The other emotion type and scene type labels are updated in the same way, which will not be repeated here.
[0129] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.
Claims
1. An intelligent analysis system based on multidimensional data portrait, characterized in that: include: The basic portrait database stores portrait data of several dimensions of users and a portrait label dataset of each user; A portrait update driving unit, which is connected to the basic portrait database, and is used to determine the update period of the portrait tag of a single user according to the portrait data of each dimension of the single user, and to determine the single trigger update timing of the portrait tag of the single user according to the relative change rate of the portrait data of any dimension in the portrait data of each dimension of the single user and the update period; Among them, the relative change rate of the numerical dimension data , For the second time point image data, To portray data at the first point in time; The relative rate of change of descriptive dimensional data , set the descriptive dimension data of a certain dimension of the single user to A1 at the first time point and to A2 at the second time point, is the number of users whose descriptive dimension data is A1 at the first time point among all users, is the number of users whose descriptive dimensional data is A1 at the first time point and whose descriptive dimensional data changes to A2 at the second time point; During the update period, if the relative change rate of the portrait data in any dimension is greater than the preset relative change rate, it is determined that the portrait data of the single user meets the data condition of the single trigger update opportunity; A portrait analysis unit, which is connected to the basic portrait database and the portrait update driving unit respectively, and is used to adjust the update cycle according to the triggering frequency of a single user portrait tag; A portrait data dynamic analysis unit, which is connected to the basic portrait database and is used to calculate the data fluctuation value of the portrait data of each dimension of a single user, determine the contribution degree of the portrait data to the portrait label of the single user according to the data fluctuation value, and determine whether to adjust the collection dimension of the portrait data of the single user in the basic portrait database according to the contribution degree; Wherein, the portrait data dynamic analysis unit calculates the data fluctuation value of the portrait data of a single dimension of a single user according to the standard deviation of the portrait data of a single dimension of a single user within a set time period; The data fluctuation value of the numerical dimension data , is the portrait data in a single dimension. is the mean of the portrait data in a single dimension, The total number of portrait data under a single dimension within a set time period; The data fluctuation value of the descriptive dimensional data , The number of changes in the profile data of a single user in a single dimension within each preset time period. It is the average of all changes in the portrait data of a single user in a single dimension within each preset time period. is the total number of all preset time periods, where the preset time periods are determined according to the update cycle; The base weight of the portrait data of the corresponding dimension is adjusted according to the calculated data fluctuation value, and the contribution of each dimension to the portrait label of a single user is determined according to the adjusted weight of each dimension; the portrait data dynamic analysis unit is provided with a weight adjustment rule, including: If the data fluctuation value is less than the first fluctuation threshold, the contribution degree is the sum of the base weight and the weight increment; If the data fluctuation value is greater than the first fluctuation threshold and less than the second fluctuation threshold, the contribution degree is the reference weight; If the data fluctuation value is greater than the second fluctuation threshold, the contribution degree is the difference between the reference weight and the weight increment; A portrait execution unit is respectively connected to the basic portrait database, the portrait update driving unit and the portrait data dynamic analysis unit, and is used to execute user portraits according to the update cycle, the single trigger update timing and the collection dimension to generate the portrait label data set corresponding to each user.
2. The intelligent analysis system based on multidimensional data portrait according to claim 1, characterized in that: The portrait update driving unit determines the update period of the single user portrait label according to the change frequency of the portrait data of each dimension of the single user. The change frequency is inversely correlated with the period length of the update period.
3. The intelligent analysis system based on multidimensional data portrait according to claim 2 is characterized in that: The portrait analysis unit adjusts the update period according to the triggering frequency of each portrait tag of a single user, including: If the trigger frequencies of the portrait tags of a single user are all less than the first set threshold, it is determined to increase the period length of the update period; If the trigger frequency of at least one portrait tag among the portrait tags of a single user is greater than or equal to the first set threshold, it is determined to reduce the period length of the update period.
4. The intelligent analysis system based on multidimensional data portrait according to claim 1 is characterized in that: The portrait update driving unit determines whether the portrait data of a single user meets the time condition of a single trigger update opportunity according to the interval between the judgment result of the data condition and the last portrait tag update, including: If the portrait data of a single user is currently judged to meet the data conditions of a single trigger update timing and the interval between the judgment result of the data conditions and the last portrait tag update is longer than the preset waiting interval, the portrait data of a single user is judged to meet the time conditions of a single trigger update timing.
5. The intelligent analysis system based on multidimensional data portrait according to claim 1 is characterized in that: The portrait data dynamic analysis unit determines whether to adjust the collection dimension of the portrait data of a single user in the basic portrait database according to the contribution degree, including: If the contribution degree is greater than the contribution degree threshold, retain the collection dimension and continue to collect the portrait data of the collection dimension; If the contribution degree is less than or equal to the contribution degree threshold, remove the collection dimension and stop collecting the portrait data of the collection dimension.
6. The intelligent analysis system based on multidimensional data portrait according to claim 1 is characterized in that: The portrait execution unit executes user portrait according to the update cycle, the single trigger update timing, and the collection dimension to generate the portrait label data set corresponding to each user, including: The portrait execution unit re-collects the portrait data under the corresponding dimension according to the collection dimension based on the update cycle or the single trigger update timing and stores it in the basic portrait database; The basic portrait database generates the updated portrait label data set corresponding to the user based on the update instruction of the portrait execution unit and the portrait data.
Citation Information
Patent Citations
User portrait label evaluation method
CN115062789A
Fault detection method and electronic equipment
CN111371938A
User portrait updating method and device, equipment and medium
CN114331486A
