App user portrait construction method and system based on multi-scene fusion
By performing correlation analysis and similarity calculation on sample data from multiple scenarios, representative tags in user profiles are identified, solving the problem of neglecting scenario differences and correlations in traditional user profile construction, and improving the accuracy and representativeness of user profiles.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU KEMAN TECHNOLOGY CO LTD
- Filing Date
- 2026-06-18
- Publication Date
- 2026-07-21
AI Technical Summary
Traditional user profiling methods ignore the differences and correlations of tags in different scenarios, resulting in inaccurate user profiles and a lack of representativeness in the extracted user features.
By performing correlation analysis on sample data from multiple scenarios, the itemsets of each scenario are identified, and the similarity and unique labels between the target itemset and candidate itemsets are calculated. The scenario differentiation strength and feature credibility are determined, and the differentiation ability of the labels in different scenarios is comprehensively evaluated. Representative labels are selected as user profile features.
It significantly improves the accuracy of user profile construction and the representativeness of feature labels, overcomes the shortcomings of traditional methods that rely on fixed single scenarios, and realizes the scenario-based decomposition and feature extraction of user consumption behavior.
Smart Images
Figure CN122434580A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data analysis technology, and in particular to a method and system for constructing APP user profiles based on multi-scenario fusion. Background Technology
[0002] Every product has a specific target audience. Manufacturers analyze the characteristics of this audience to discover common traits, thereby identifying relevant demographic features. They can then promote their products to individuals with these features, achieving precise marketing and personalized service recommendations. Understanding user preferences through user profiling is a crucial step in this process.
[0003] Traditional user profiling typically relies on a single, fixed scenario, analyzing user preferences within that scenario to create a user profile. Extracting user features is a core step in user profiling. Traditional methods usually involve a holistic analysis of the provided data, using feature extraction algorithms to obtain key user profile tags. However, this approach overlooks the differences in tags across different scenarios, making it impossible to determine the correlation between tags. Consequently, the resulting user profiles are inaccurate, and the extracted user features lack representativeness. Summary of the Invention
[0004] To address the aforementioned technical issues, the purpose of this application is to provide a method and system for constructing APP user profiles based on multi-scenario fusion. The specific technical solution adopted is as follows: Firstly, a method for constructing APP user profiles based on multi-scenario fusion is provided, the method comprising: Association analysis was performed on sample data from multiple scenarios to obtain itemsets for each scenario; sample data from multiple scenarios were obtained by segmenting merchant transaction data into scenarios. Based on the similarity between the target itemset in the scene to be analyzed and the candidate itemsets in other scenes, the corresponding matching itemset pairs are obtained, and the common tags of the matching itemset pairs and the unique tags of the target itemset are determined. The discrimination strength of the matching itemset is calculated based on the frequency of occurrence of the unique labels of the corresponding target itemset in different scenarios. The scene discrimination strength of the matching itemset is determined based on the difference in the frequency of occurrence of the common labels of the matching itemset. The feature credibility of the matching itemset is determined based on the proportion of sample data in the scene corresponding to the matching itemset. For each target itemset in the scenario to be analyzed, the scenario discrimination coefficient of the target itemset is calculated based on the discrimination strength of multiple matching itemset pairs corresponding to the target itemset, the scenario discrimination strength, and the feature credibility. The labels in the preferred itemset are used as user profile features in the scenario to be analyzed. The preferred itemset indicates the itemset whose scenario discrimination coefficient meets the preset conditions.
[0005] Optionally, association analysis can be performed on sample data from multiple scenarios to obtain itemsets for each scenario, including: Based on preset scenario segmentation rules, the merchant's transaction data is classified into multiple data subsets; each data subset corresponds to a scenario. Association analysis algorithms are used to perform association analysis on sample data in each scenario to obtain at least one maximum frequent itemset in each scenario; where the itemset includes the maximum frequent itemset.
[0006] Optionally, based on the similarity between the target itemset in the scene to be analyzed and candidate itemsets in other scenes, corresponding matching itemset pairs are obtained, and the common tags of the matching itemset pairs and the unique tags of the target itemset are determined, including: Calculate the similarity between the target itemset in the scenario to be analyzed and multiple candidate itemsets in each other scenario, and select the candidate itemset with the highest similarity and greater than the preset filtering threshold. Associate the target itemset with the selected candidate itemset to obtain a matching itemset pair. The intersection of the two corresponding itemsets of a matching itemset pair is determined as the common label of the matching itemset pair; The unique labels of the target itemset are determined by the difference between the common labels of the target itemset and the matching itemset pair.
[0007] Optionally, the discriminative strength of the matching itemset pair is calculated based on the frequency of occurrence of its unique labels in different scenarios, and the scenario discriminative strength of the matching itemset pair is determined based on the difference in the frequency of occurrence of its common labels. The feature reliability of the matching itemset pair is determined based on the proportion of sample data in the scenario corresponding to the matching itemset pair, including: Obtain the target itemset in the scenario to be analyzed, and obtain multiple matching itemset pairs formed by the target itemset and candidate itemsets in multiple other scenarios; For each matching item pair, obtain the unique label in the target item set, and count the frequency of occurrence of the unique label in the unique labels of multiple matching item pairs. Determine the distinguishing strength of the matching item pair based on the exponential decay value of the occurrence frequency. Obtain the common label of the matching itemset pair, calculate the first occurrence frequency of the common label in the sample data of the scene to be analyzed corresponding to the matching itemset pair, and the second occurrence frequency of the common label in the sample data of other scenes corresponding to the matching itemset pair, and determine the scene discrimination strength of the matching itemset pair based on the difference between the first occurrence frequency and the second occurrence frequency. The conversion rate of a scenario is determined by the ratio between the number of sample data in each scenario and the number of sample data across multiple scenarios. Determine the first scenario conversion rate of the matching itemset for the corresponding scenario to be analyzed, and determine the second scenario conversion rate of the matching itemset for the corresponding other scenarios. Calculate the first sum of the conversion rates of the first scenario and the conversion rates of the second scenario, and calculate the second sum of the first frequency of occurrence and the second frequency of occurrence; The feature confidence of the matching itemset pair is determined by the product of the first sum and the second sum.
[0008] Optionally, for each target itemset in the scenario to be analyzed, the scenario discrimination coefficient of the target itemset is calculated based on the discrimination strength of multiple matching itemset pairs corresponding to the target itemset, the scenario discrimination strength, and the feature credibility. The labels in the preferred itemset are then used as user profile features in the scenario to be analyzed, including: The overall discrimination strength of a matching itemset is obtained by multiplying the discrimination strength of each matching itemset pair by the scene discrimination strength of that matching itemset pair. The weighted discrimination coefficient of each matching itemset is obtained by calculating the ratio of the feature confidence of each matching itemset to the similarity of that matching itemset, and by multiplying the overall discrimination strength of the matching itemset with the ratio. For each target itemset in the scenario to be analyzed, obtain multiple matching itemset pairs corresponding to that target itemset. Determine the scenario discrimination coefficient of the target itemset based on the sum of the weighted discrimination coefficients of the multiple matching itemset pairs. For multiple target itemsets in the scenario to be analyzed, the target itemsets with a scenario discrimination coefficient greater than a preset threshold are determined as the preferred itemsets for the scenario to be analyzed, and the tags in the preferred itemsets are used as user profile features in the scenario to be analyzed.
[0009] Secondly, a system for constructing APP user profiles based on multi-scenario fusion is provided, the system comprising: The analysis module is used to perform correlation analysis on sample data from multiple scenarios to obtain itemsets for each scenario; the sample data from multiple scenarios is obtained by dividing the merchant's transaction data into scenarios. The determination module is used to obtain the corresponding matching item set pairs based on the similarity between the target item set in the scene to be analyzed and the candidate item sets in other scenes, and to determine the common tags of the matching item set pairs and the unique tags of the target item set. The first calculation module is used to calculate the discrimination strength of the matching itemset pair based on the frequency of occurrence of the unique labels of the corresponding target itemset in different scenarios, determine the scenario discrimination strength of the matching itemset pair based on the difference in the frequency of occurrence of the common labels of the matching itemset pair, and determine the feature credibility of the matching itemset pair based on the proportion of sample data in the scenario corresponding to the matching itemset pair. The second calculation module is used to calculate the scene discrimination coefficient of each target itemset in the scene to be analyzed based on the discrimination strength, scene discrimination strength and feature credibility of the multiple matching itemset pairs corresponding to the target itemset, and to use the labels in the preferred itemset as user profile features in the scene to be analyzed; the preferred itemset indicates the itemset whose scene discrimination coefficient meets the preset conditions.
[0010] Optionally, the analysis module is also used for: Based on preset scenario segmentation rules, the merchant's transaction data is classified into multiple data subsets; each data subset corresponds to a scenario. Association analysis algorithms are used to perform association analysis on sample data in each scenario to obtain at least one maximum frequent itemset in each scenario; where the itemset includes the maximum frequent itemset.
[0011] Optionally, the determining module is also used for: Calculate the similarity between the target itemset in the scenario to be analyzed and multiple candidate itemsets in each other scenario, and select the candidate itemset with the highest similarity and greater than the preset filtering threshold. Associate the target itemset with the selected candidate itemset to obtain a matching itemset pair. The intersection of the two corresponding itemsets of a matching itemset pair is determined as the common label of the matching itemset pair; The unique labels of the target itemset are determined by the difference between the common labels of the target itemset and the matching itemset pair.
[0012] Optionally, the first computing module is also used for: Obtain the target itemset in the scenario to be analyzed, and obtain multiple matching itemset pairs formed by the target itemset and candidate itemsets in multiple other scenarios; For each matching item pair, obtain the unique label in the target item set, and count the frequency of occurrence of the unique label in the unique labels of multiple matching item pairs. Determine the distinguishing strength of the matching item pair based on the exponential decay value of the occurrence frequency. Obtain the common label of the matching itemset pair, calculate the first occurrence frequency of the common label in the sample data of the scene to be analyzed corresponding to the matching itemset pair, and the second occurrence frequency of the common label in the sample data of other scenes corresponding to the matching itemset pair, and determine the scene discrimination strength of the matching itemset pair based on the difference between the first occurrence frequency and the second occurrence frequency. The conversion rate of a scenario is determined by the ratio between the number of sample data in each scenario and the number of sample data across multiple scenarios. Determine the first scenario conversion rate of the matching itemset for the corresponding scenario to be analyzed, and determine the second scenario conversion rate of the matching itemset for the corresponding other scenarios. Calculate the first sum of the conversion rates of the first scenario and the conversion rates of the second scenario, and calculate the second sum of the first frequency of occurrence and the second frequency of occurrence; The feature confidence of the matching itemset pair is determined by the product of the first sum and the second sum.
[0013] Optionally, the second computing module is also used for: The overall discrimination strength of a matching itemset is obtained by multiplying the discrimination strength of each matching itemset pair by the scene discrimination strength of that matching itemset pair. The weighted discrimination coefficient of each matching itemset is obtained by calculating the ratio of the feature confidence of each matching itemset to the similarity of that matching itemset, and by multiplying the overall discrimination strength of the matching itemset with the ratio. For each target itemset in the scenario to be analyzed, obtain multiple matching itemset pairs corresponding to that target itemset. Determine the scenario discrimination coefficient of the target itemset based on the sum of the weighted discrimination coefficients of the multiple matching itemset pairs. For multiple target itemsets in the scenario to be analyzed, the target itemsets with a scenario discrimination coefficient greater than a preset threshold are determined as the preferred itemsets for the scenario to be analyzed, and the tags in the preferred itemsets are used as user profile features in the scenario to be analyzed.
[0014] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this application.
[0015] This application offers the following advantages: By segmenting merchant transaction data into scenarios, sample data from multiple scenarios is obtained. Correlation analysis of the sample data in each scenario yields itemsets for each scenario, enabling scenario-based decomposition of user consumption behavior. By calculating the similarity between the target itemset in the analyzed scenario and candidate itemsets in other scenarios, matching itemset pairs are obtained. The common tags of the matching itemset pairs and the unique tags of the target itemset are determined, effectively identifying the correlation and differences of tags in different scenarios. The discrimination strength is calculated based on the frequency of occurrence of the unique tags of the target itemset corresponding to the matching itemset pair in different scenarios. The scenario discrimination strength is determined based on the difference in the frequency of occurrence of the common tags of the matching itemset pairs. The feature credibility is determined based on the proportion of sample data in the scenario corresponding to the matching itemset pair, thus comprehensively evaluating the discrimination ability of tags in different scenarios from multiple dimensions. For each target itemset in the analyzed scenario, the scenario discrimination coefficient of the target itemset is calculated based on the discrimination strength of multiple matching itemset pairs corresponding to that target itemset, the scenario discrimination strength, and the feature credibility. Tags in itemsets whose scenario discrimination coefficients meet preset conditions are used as user profile features in the analyzed scenario. It overcomes the shortcomings of traditional user profile construction, which fixes a single scenario and ignores the differences and correlations of tags in different scenarios. By extracting representative scenario features through multi-scenario fusion analysis, it significantly improves the accuracy of user profile construction and the representativeness of feature tags. Attached Figure Description
[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart of an APP user profile construction method based on multi-scenario fusion in one embodiment; Figure 2 This is a schematic diagram of the structure of an APP user profile construction system based on multi-scenario fusion in one embodiment; Figure 3 This is a schematic diagram of the structure of an electronic device in one embodiment. Detailed Implementation
[0018] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method and system for constructing APP user profiles based on multi-scenario fusion proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.
[0020] The following section, with reference to the accompanying drawings, details a specific scheme for an APP user profile construction method based on multi-scenario fusion provided in this application. For example... Figure 1 As shown, the method includes: S11. Perform association analysis on sample data in multiple scenarios to obtain itemsets for each scenario.
[0021] The sample data for multiple scenarios were obtained by dividing the merchants' transaction data into scenarios.
[0022] The analysis focuses on a specific merchant. User consumption data for that merchant is obtained (through historical transaction records), such as purchase amount, average price, product attributes, and category. This data can be divided into consumer-related data and product-related data. The explanation of product-related data is as follows: For a clothing store, there are various styles of clothing, with tags such as color, size, short sleeves, and outerwear. The same product may have multiple tags; each product is considered a sample. Therefore, a purchase record consists of multiple tags.
[0023] The process involves acquiring historical consumption data from merchants to create a sample set. This sample set is then categorized based on the merchants' user characteristics, resulting in sample data across multiple scenarios. For example, consumption scenarios can be segmented based on consumer attributes such as age and gender, or specifically based on the consumption characteristics of young women. It's important to note that the above is just an example; in real-world scenarios, other attributes requiring further research can be used for differentiation, such as time (based on season, weekday status, and consumption time). Different merchants require different classification methods; for instance, restaurant merchants rely more heavily on time-based classification.
[0024] Because the user classifications mentioned above are influenced by the products, different groups hold varying degrees of importance for merchants. Therefore, this merchant has primary sales targets; for example, younger users often buy sneakers and skate shoes, while older users prefer canvas shoes. This merchant primarily serves younger consumers with purchasing power, thus sales targets differ across different scenarios.
[0025] Choose one scenario and calculate its conversion rate. A higher conversion rate indicates that the scenario has a high conversion characteristic rate for the merchant. The scenario conversion characteristic rate reflects the importance and value of the scenario to the merchant. The scenario conversion rate is obtained by taking the proportion of sample data from each scenario out of the total sample data from all scenarios, denoted as . This is used to quantify the value of a scenario, thereby identifying more important and frequent scenarios for merchants, and providing scenario importance weights for subsequent extraction of user profile features. For example, if a merchant's data shows that most consumers are young users, then the merchant should provide more product recommendations to this user group; conversely, if such scenarios account for a small percentage of the merchant's data, it indicates that such scenarios are not important to the merchant and are not worth recommending.
[0026] In one embodiment, association analysis is performed on sample data from multiple scenarios to obtain itemsets for each scenario, including: Based on preset scenario segmentation rules, the merchant's transaction data is classified into multiple data subsets; each data subset corresponds to a scenario. Association analysis algorithms are used to perform association analysis on sample data in each scenario to obtain at least one maximum frequent itemset in each scenario; where the itemset includes the maximum frequent itemset.
[0027] The process involves acquiring relevant historical consumption data from merchants to obtain a sample set. Based on the merchants' user characteristics, this sample set is categorized to obtain a data subset for each scenario. Each data subset includes sample data from each scenario. A scenario is selected, and the TF-tree algorithm (an association analysis algorithm) or other algorithms are used to perform association analysis on the sample data for each scenario. This yields at least one maximum frequent itemset for that scenario. The maximum frequent itemset reflects the main characteristics of the consumption records in that scenario. Subsequent steps in this application will refer to the maximum frequent itemset simply as an itemset, and the labels within the itemset indicate the association.
[0028] S12. Based on the similarity between the target itemset in the scene to be analyzed and the candidate itemsets in other scenes, obtain the corresponding matching itemset pairs, and determine the common labels of the matching itemset pairs and the unique labels of the target itemset.
[0029] In one embodiment, based on the similarity between the target itemset in the scenario to be analyzed and candidate itemsets in other scenarios, corresponding matching itemset pairs are obtained, and the common tags of the matching itemset pairs and the unique tags of the target itemset are determined, including: Calculate the similarity between the target itemset in the scenario to be analyzed and multiple candidate itemsets in each other scenario, and select the candidate itemset with the highest similarity and greater than the preset filtering threshold. Associate the target itemset with the selected candidate itemset to obtain a matching itemset pair. The intersection of the two corresponding itemsets of a matching itemset pair is determined as the common label of the matching itemset pair; The unique labels of the target itemset are determined by the difference between the common labels of the target itemset and the matching itemset pair.
[0030] Since scene differentiation is based on pre-selected human choices, different profile features of people in different scenes are obtained. However, not all of these features can be used to construct user profiles. For example, although the obtained sample labels reflect the common characteristics of consumers in that scene, these features do not change when the scene changes. Therefore, such profile features cannot be used as the main characteristics of users. Thus, the higher the feature similarity between two scenes, the weaker the distinguishing ability of that feature.
[0031] Obtain the j-th target itemset in the i-th scenario to be analyzed. Compare it with the l-th candidate set in the k-th other scenario. Comparison, calculated using intersection-union ratio. and similarity between If two itemsets are completely identical, their similarity is 1. Higher similarity indicates greater consistency across different scenarios, reflecting shared consumption tags within the itemsets. This applies to the j-th target itemset in the i-th scenario to be analyzed. Iterate through all candidate sets in the kth other scene, and determine the candidate set with the highest similarity in the kth other scene that is greater than the preset filtering threshold as the one with the highest similarity. A matching itemset will By associating with this candidate itemset, matching itemset pairs are obtained. The candidate itemset is a matching itemset.
[0032] It should be noted that the target itemset There is at most one matching itemset with each other scenario, namely the target itemset. This corresponds to multiple matching itemsets, i.e., target itemsets. This corresponds to multiple matching itemset pairs. If no matching itemset exists in other scenarios... Candidate itemsets with a similarity greater than a preset filtering threshold are then considered the target itemset. There is no matching itemset for this other scenario.
[0033] A matching itemset pair refers to a pairing relationship between two itemsets. It is a correspondence formed by the target itemset in the scenario to be analyzed and the candidate itemsets in other scenarios that match it, rather than merging two itemsets into one set.
[0034] and similarity The calculation formula is: ; in, This represents the intersection-union ratio of two matching itemsets.
[0035] Consumption-related tags common to different scenarios are obtained through threshold filtering. A filtering threshold is set. The screening threshold can be 0.5, 0.6, or 0.7. If... If the value is greater than the threshold, the two itemsets are considered a correct match. and A matching itemset pair is formed; otherwise, an incorrect match is made. and They do not form a matching itemset pair.
[0036] For a pair of matching itemsets, calculate the common label of the pair of matching itemsets. and Common tags of the matching itemset pairs The calculation formula is: .Record and The similarity and shared consumption tags indicate the itemset The tags mentioned above exist in different scenarios, so the common tags mentioned above cannot be used as unique features of a particular scenario. However, if a scenario does not have a common tag, it means that the tag is more likely to be a unique feature of that scenario, and thus can be used to build a user profile.
[0037] and In the constructed matching itemset pairs, the target itemset Unique tags The calculation method is as follows: .
[0038] The itemset differentiation process primarily relies on the accuracy of the intersection-union ratio (IUR) calculation and the correlation between different labels within an itemset. However, in actual processing, while labels within an itemset may indicate frequent co-occurrence, they don't necessarily possess a correlation. For example, when differentiating scenarios by age group, younger age groups might purchase {down jacket, dark-colored} clothing, while older age groups might have different labels. Clothing. Clearly, the only potentially distinguishing element between the two most frequent itemsets is clothing color. Whether or not it's a down jacket is irrelevant to the age segmentation scenario mentioned above; therefore, the "down jacket" tag lacks scenario-distinguishing capabilities. However, while both "down jacket" and "color" form a most frequent itemset, and the two itemsets are not identical, some tags within these itemsets cannot actually serve as scenario-distinguishing labels. Therefore, it's necessary to consider the overlap of labels within each scenario.
[0039] S13. Calculate the distinguishing strength of the matching itemset pair based on the frequency of occurrence of the unique labels of the target itemset in different scenarios, determine the scene distinguishing strength of the matching itemset pair based on the difference in the frequency of occurrence of the common labels of the matching itemset pair, and determine the feature credibility of the matching itemset pair based on the proportion of sample data in the scene corresponding to the matching itemset pair.
[0040] In one embodiment, the discriminative strength of a matching itemset pair is calculated based on the frequency of occurrence of its unique labels in different scenarios. The scenario discriminative strength of the matching itemset pair is determined based on the difference in the frequency of occurrence of its common labels. The feature reliability of the matching itemset pair is determined based on the proportion of sample data in the scenario corresponding to the matching itemset pair, including: Obtain the target itemset in the scenario to be analyzed, and obtain multiple matching itemset pairs formed by the target itemset and candidate itemsets in multiple other scenarios; For each matching item pair, obtain the unique label in the target item set, and count the frequency of occurrence of the unique label in the unique labels of multiple matching item pairs. Determine the distinguishing strength of the matching item pair based on the exponential decay value of the occurrence frequency. Obtain the common label of the matching itemset pair, calculate the first occurrence frequency of the common label in the sample data of the scene to be analyzed corresponding to the matching itemset pair, and the second occurrence frequency of the common label in the sample data of other scenes corresponding to the matching itemset pair, and determine the scene discrimination strength of the matching itemset pair based on the difference between the first occurrence frequency and the second occurrence frequency. The conversion rate of a scenario is determined by the ratio between the number of sample data in each scenario and the number of sample data across multiple scenarios. Determine the first scenario conversion rate of the matching itemset for the corresponding scenario to be analyzed, and determine the second scenario conversion rate of the matching itemset for the corresponding other scenarios. Calculate the first sum of the conversion rates of the first scenario and the conversion rates of the second scenario, and calculate the second sum of the first frequency of occurrence and the second frequency of occurrence; The feature confidence of the matching itemset pair is determined by the product of the first sum and the second sum.
[0041] The distinguishability strength characterizes the uniqueness of the unique label in the matched itemset pair. The larger the value, the less likely the unique label appears in other scenarios, and the more representative it is of the unique features of the current scenario; the smaller the value, the more likely the unique label appears frequently in other scenarios, and the weaker the distinguishability.
[0042] Scene discrimination strength characterizes the degree of difference in the frequency of occurrence of common labels in matching itemsets across different scenes. The larger the value, the greater the difference in the frequency of occurrence of the common label in the two scenes, and the better it can be used to distinguish between different scenes; the smaller the value, the closer the frequency of occurrence of the common label in the two scenes, and the weaker the discrimination ability.
[0043] Feature credibility characterizes the reliability of feature analysis between two itemsets in a matching itemset pair. This value comprehensively considers the sample size ratio of the two scenarios (scenario conversion rate) and the frequency of occurrence of the two itemsets in their respective scenarios. The more samples and the higher the frequency of itemset occurrence, the higher the feature credibility.
[0044] Select a target itemset from the scenario to be analyzed. Obtain multiple matching itemset pairs between the target itemset and candidate itemsets from other scenarios. For each matching itemset pair, extract the unique tags from the target itemset to obtain a set of unique tags from the multiple matching itemset pairs. For each matching itemset pair, obtain each unique tag in the target itemset and calculate the value of that unique tag in the unique tag set. Frequency of occurrence in The frequency of occurrence reflects whether the scene features (represented by labels) obtained in the current scene also exist in other scenes. If the analyzed label is a label of the scene feature, then it exists in the feature label set in the comparison of other scenes. The fewer times it exists, the better the current scene can be distinguished from other scenes by that label. Conversely, if the label is not representative of the scene, then the obtained label should exist in other scenes, thus appearing more frequently in the unique label set.
[0045] The above analysis was performed on each unique tag in the target item set, and the frequency of occurrence of each unique tag was recorded. The frequencies are added together, and the discriminative strength of the matching itemset is determined based on the exponential decay of the sum of the frequencies. and Discriminating strength of the matching itemset pairs The calculation formula is: exp() is an exponential function with the natural constant e as its base.
[0046] Different itemsets in different scenarios can still reflect the unique characteristics of that scenario. For example, scenarios are divided by season; people typically buy clothing for a specific season, but they might also buy clothing for other seasons. Taking advantage of the limited sample data in each scenario, if the frequency of itemsets remains consistent across different scenarios, it indicates that there is no significant change in the itemsets after changing scenarios, and therefore the feature is scenario-independent. However, if the frequency of itemsets changes significantly between different scenarios when the scenario changes, then even if there are identical itemsets, these itemsets can still serve as a feature specific to that scenario.
[0047] Therefore, for the common label of the matching itemset pairs Obtain shared tags The first frequency of occurrence in the sample data of the i-th scenario to be analyzed Specifically, the sample data of the i-th scenario to be analyzed simultaneously contains Number of samples for all labels Obtain the total number of sample data for the i-th scenario to be analyzed. ,calculate and The ratio of the two values yields the first frequency of occurrence. Similarly, calculate the second frequency of occurrence of the shared label in the sample data of the kth other scene. , and Scene discrimination strength of the matching itemset pairs The calculation formula is: norm() is the normalization formula used to normalize the data. Normalize to the [0,1] interval, such as Min-Max normalization, Sigmoid normalization, etc.
[0048] The above analysis of intensity is easily affected by the number of samples. If the number of consumption records in a scenario is small, the stability of feature extraction for that scenario will be worse. Even slight changes in the number of samples may result in significant differences in the user profiles constructed, making the analysis of scenario tags unreliable. Therefore, the scenario conversion rate and intensity in each scenario are used as the reliability of feature calculation.
[0049] Determine the first scenario conversion rate for the i-th scenario to be analyzed. Determine the conversion rate of the second scenario for the kth other scenario. The scene conversion rate is obtained by comparing the number of sample data in that scene with the total number of samples across multiple scenes. and Feature credibility of the matching itemset pairs The calculation formula is: .
[0050] S14. For each target itemset in the scenario to be analyzed, calculate the scenario discrimination coefficient of the target itemset based on the discrimination strength, scenario discrimination strength and feature credibility of the multiple matching itemset pairs corresponding to the target itemset, and use the labels in the preferred itemset as user profile features in the scenario to be analyzed.
[0051] Among them, the preferred itemset indicates the itemset whose scene discrimination coefficient meets the preset conditions.
[0052] In one embodiment, for each target itemset in the scenario to be analyzed, the scenario discrimination coefficient of the target itemset is calculated based on the discrimination strength of multiple matching itemset pairs corresponding to the target itemset, the scenario discrimination strength, and the feature credibility. The labels in the preferred itemset are then used as user profile features in the scenario to be analyzed, including: The overall discrimination strength of a matching itemset is obtained by multiplying the discrimination strength of each matching itemset pair by the scene discrimination strength of that matching itemset pair. The weighted discrimination coefficient of each matching itemset is obtained by calculating the ratio of the feature confidence of each matching itemset to the similarity of that matching itemset, and by multiplying the overall discrimination strength of the matching itemset with the ratio. For each target itemset in the scenario to be analyzed, obtain multiple matching itemset pairs corresponding to that target itemset. Determine the scenario discrimination coefficient of the target itemset based on the sum of the weighted discrimination coefficients of the multiple matching itemset pairs. For multiple target itemsets in the scenario to be analyzed, the target itemsets with a scenario discrimination coefficient greater than a preset threshold are determined as the preferred itemsets for the scenario to be analyzed, and the tags in the preferred itemsets are used as user profile features in the scenario to be analyzed.
[0053] The overall distinguishing strength characterizes the overall distinguishing ability of the matching itemset pair, comprehensively reflecting the uniqueness of the unique label and the scene difference of the shared label. The larger the value, the stronger the distinguishing ability of the matching itemset pair.
[0054] Based on the discriminative strength of matching itemset pairs Distinguishing Intensity from Scene The product of these factors yields the overall discriminative strength of the matching itemset pair. and The overall discriminative strength of the matching itemset pairs The calculation formula is: .
[0055] Target Itemset Scene discrimination coefficient The calculation formula is: ; Where k is the k-th other scenario, and K is the number of other scenarios. for and The feature credibility of the matching itemset pairs constituted. for and similarity, for and The overall distinguishing strength of the matching itemset pairs.
[0056] The greater the scene discrimination capability of the target itemset, the better the feature labels within the target itemset can reflect the characteristics of that scene. Therefore, the target itemset is used as the user profile feature for that scene. A preset threshold H is set (H can be 0.5 or 0.6), and target itemsets exceeding the preset threshold are filtered to obtain a representative preferred itemset for that scene. The labels in the preferred itemset are then used as the user profile features for the scene to be analyzed.
[0057] For example, merchants can obtain user profile features through the APP in the following ways: Merchants upload transaction-related data, and for the relevant tags in the transaction data, merchants manually select some tags as objects to be analyzed (the tags can be changed), or use feature selection algorithms (such as principal component analysis to select the tag corresponding to the largest principal component, which automatically realizes the tag selection). Based on the selected tags, the scenarios are divided. The APP uses the above algorithm to obtain user profile features in each scenario, thereby realizing the construction of user profiles.
[0058] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.
[0059] This application also provides an APP user profile construction system based on multi-scenario fusion, such as... Figure 2 As shown, the system includes: Analysis module 21 is used to perform correlation analysis on sample data in multiple scenarios to obtain itemsets for each scenario; the sample data in multiple scenarios is obtained by dividing the merchant's transaction data into scenarios; The determination module 22 is used to obtain the corresponding matching item set pairs based on the similarity between the target item set in the scene to be analyzed and the candidate item sets in other scenes, and to determine the common tags of the matching item set pairs and the unique tags of the target item set. The first calculation module 23 is used to calculate the discrimination strength of the matching itemset pair based on the frequency of occurrence of the unique labels of the target itemset corresponding to the matching itemset pair in different scenarios, and to determine the scenario discrimination strength of the matching itemset pair based on the difference in the frequency of occurrence of the common labels of the matching itemset pair, and to determine the feature credibility of the matching itemset pair based on the proportion of sample data in the scenario corresponding to the matching itemset pair. The second calculation module 24 is used to calculate the scene discrimination coefficient of each target itemset in the scene to be analyzed based on the discrimination strength, scene discrimination strength and feature credibility of the multiple matching itemset pairs corresponding to the target itemset, and to use the labels in the preferred itemset as user profile features in the scene to be analyzed; the preferred itemset indicates the itemset whose scene discrimination coefficient meets the preset conditions.
[0060] Optionally, the analysis module 21 is also used for: Based on preset scenario segmentation rules, the merchant's transaction data is classified into multiple data subsets; each data subset corresponds to a scenario. Association analysis algorithms are used to perform association analysis on sample data in each scenario to obtain at least one maximum frequent itemset in each scenario; where the itemset includes the maximum frequent itemset.
[0061] Optionally, the determining module 22 is also used for: Calculate the similarity between the target itemset in the scenario to be analyzed and multiple candidate itemsets in each other scenario, and select the candidate itemset with the highest similarity and greater than the preset filtering threshold. Associate the target itemset with the selected candidate itemset to obtain a matching itemset pair. The intersection of the two corresponding itemsets of a matching itemset pair is determined as the common label of the matching itemset pair; The unique labels of the target itemset are determined by the difference between the common labels of the target itemset and the matching itemset pair.
[0062] Optionally, the first calculation module 23 is also used for: Obtain the target itemset in the scenario to be analyzed, and obtain multiple matching itemset pairs formed by the target itemset and candidate itemsets in multiple other scenarios; For each matching item pair, obtain the unique label in the target item set, and count the frequency of occurrence of the unique label in the unique labels of multiple matching item pairs. Determine the distinguishing strength of the matching item pair based on the exponential decay value of the occurrence frequency. Obtain the common label of the matching itemset pair, calculate the first occurrence frequency of the common label in the sample data of the scene to be analyzed corresponding to the matching itemset pair, and the second occurrence frequency of the common label in the sample data of other scenes corresponding to the matching itemset pair, and determine the scene discrimination strength of the matching itemset pair based on the difference between the first occurrence frequency and the second occurrence frequency. The conversion rate of a scenario is determined by the ratio between the number of sample data in each scenario and the number of sample data across multiple scenarios. Determine the first scenario conversion rate of the matching itemset for the corresponding scenario to be analyzed, and determine the second scenario conversion rate of the matching itemset for the corresponding other scenarios. Calculate the first sum of the conversion rates of the first scenario and the conversion rates of the second scenario, and calculate the second sum of the first frequency of occurrence and the second frequency of occurrence; The feature confidence of the matching itemset pair is determined by the product of the first sum and the second sum.
[0063] Optionally, the second calculation module 24 is also used for: The overall discrimination strength of a matching itemset is obtained by multiplying the discrimination strength of each matching itemset pair by the scene discrimination strength of that matching itemset pair. The weighted discrimination coefficient of each matching itemset is obtained by calculating the ratio of the feature confidence of each matching itemset to the similarity of that matching itemset, and by multiplying the overall discrimination strength of the matching itemset with the ratio. For each target itemset in the scenario to be analyzed, obtain multiple matching itemset pairs corresponding to that target itemset. Determine the scenario discrimination coefficient of the target itemset based on the sum of the weighted discrimination coefficients of the multiple matching itemset pairs. For multiple target itemsets in the scenario to be analyzed, the target itemsets with a scenario discrimination coefficient greater than a preset threshold are determined as the preferred itemsets for the scenario to be analyzed, and the tags in the preferred itemsets are used as user profile features in the scenario to be analyzed.
[0064] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs.
[0065] Figure 3 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the method described in any of the above embodiments. Figure 3 The electronic device 30 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.
[0066] like Figure 3 As shown, the electronic device 30 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).
[0067] Bus 33 includes a data bus, an address bus, and a control bus.
[0068] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.
[0069] The memory 32 may also include a program tool 325 (or utility) having a set (at least one) program module 324, such program module 324 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.
[0070] The processor 31 executes various functional applications and data processing, such as the methods provided in any of the above embodiments, by running computer programs stored in the memory 32.
[0071] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed via input / output (I / O) interface 35. Furthermore, electronic device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 36. As shown, network adapter 36 communicates with other modules of electronic device 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.
[0072] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.
[0073] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in any of the above embodiments.
[0074] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.
[0075] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0076] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the method described in any of the above embodiments.
[0077] The program code for executing the computer program product of this application can be written in any combination of one or more programming languages. The program code can be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on a remote device.
[0078] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0079] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.
[0080] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
Claims
1. A method for constructing APP user profiles based on multi-scenario fusion, characterized in that, The method includes: Association analysis was performed on sample data from multiple scenarios to obtain itemsets for each scenario; sample data from multiple scenarios were obtained by segmenting merchant transaction data into scenarios. Based on the similarity between the target itemset in the scene to be analyzed and the candidate itemsets in other scenes, the corresponding matching itemset pairs are obtained, and the common tags of the matching itemset pairs and the unique tags of the target itemset are determined. The discrimination strength of the matching itemset is calculated based on the frequency of occurrence of the unique labels of the corresponding target itemset in different scenarios. The scene discrimination strength of the matching itemset is determined based on the difference in the frequency of occurrence of the common labels of the matching itemset. The feature credibility of the matching itemset is determined based on the proportion of sample data in the scene corresponding to the matching itemset. For each target itemset in the scenario to be analyzed, the scenario discrimination coefficient of the target itemset is calculated based on the discrimination strength of multiple matching itemset pairs corresponding to the target itemset, the scenario discrimination strength, and the feature credibility. The labels in the preferred itemset are used as user profile features in the scenario to be analyzed. The preferred itemset indicates the itemset whose scenario discrimination coefficient meets the preset conditions.
2. The method for constructing APP user profiles based on multi-scenario fusion as described in claim 1, characterized in that, The step involves performing correlation analysis on sample data from multiple scenarios to obtain itemsets for each scenario, including: Based on preset scenario segmentation rules, the merchant's transaction data is classified into multiple data subsets; each data subset corresponds to a scenario. Association analysis algorithms are used to perform association analysis on sample data in each scenario to obtain at least one maximum frequent itemset in each scenario; where the itemset includes the maximum frequent itemset.
3. The method for constructing APP user profiles based on multi-scenario fusion as described in claim 1, characterized in that, The process of obtaining corresponding matching itemset pairs based on the similarity between the target itemset in the scene to be analyzed and candidate itemsets in other scenes, and determining the common tags of the matching itemset pairs and the unique tags of the target itemset, includes: Calculate the similarity between the target itemset in the scenario to be analyzed and multiple candidate itemsets in each other scenario, and select the candidate itemset with the highest similarity and greater than the preset filtering threshold. Associate the target itemset with the selected candidate itemset to obtain a matching itemset pair. The intersection of the two corresponding itemsets of a matching itemset pair is determined as the common label of the matching itemset pair; The unique labels of the target itemset are determined by the difference between the common labels of the target itemset and the matching itemset pair.
4. The method for constructing APP user profiles based on multi-scenario fusion as described in claim 1, characterized in that, The process involves calculating the discriminative strength of a matching itemset pair based on the frequency of occurrence of its unique labels in different scenarios, determining the scenario discriminative strength of the matching itemset pair based on the difference in the frequency of occurrence of its common labels, and determining the feature reliability of the matching itemset pair based on the proportion of sample data in the scenario corresponding to the matching itemset pair. This includes: Obtain the target itemset in the scenario to be analyzed, and obtain multiple matching itemset pairs formed by the target itemset and candidate itemsets in multiple other scenarios; For each matching item pair, obtain the unique label in the target item set, and count the frequency of occurrence of the unique label in the unique labels of multiple matching item pairs. Determine the distinguishing strength of the matching item pair based on the exponential decay value of the occurrence frequency. Obtain the common label of the matching itemset pair, calculate the first occurrence frequency of the common label in the sample data of the scene to be analyzed corresponding to the matching itemset pair, and the second occurrence frequency of the common label in the sample data of other scenes corresponding to the matching itemset pair, and determine the scene discrimination strength of the matching itemset pair based on the difference between the first occurrence frequency and the second occurrence frequency. The conversion rate of a scenario is determined by the ratio between the number of sample data in each scenario and the number of sample data across multiple scenarios. Determine the first scenario conversion rate of the matching itemset for the corresponding scenario to be analyzed, and determine the second scenario conversion rate of the matching itemset for the corresponding other scenarios. Calculate the first sum of the conversion rates of the first scenario and the conversion rates of the second scenario, and calculate the second sum of the first frequency of occurrence and the second frequency of occurrence; The feature confidence of the matching itemset pair is determined by the product of the first sum and the second sum.
5. The method for constructing APP user profiles based on multi-scenario fusion as described in claim 1, characterized in that, For each target itemset in the scenario to be analyzed, the scenario discrimination coefficient of the target itemset is calculated based on the discrimination strength of multiple matching itemset pairs corresponding to the target itemset, the scenario discrimination strength, and the feature credibility. The tags in the preferred itemset are then used as user profile features in the scenario to be analyzed, including: The overall discrimination strength of a matching itemset is obtained by multiplying the discrimination strength of each matching itemset pair by the scene discrimination strength of that matching itemset pair. The weighted discrimination coefficient of each matching itemset is obtained by calculating the ratio of the feature confidence of each matching itemset to the similarity of that matching itemset, and by multiplying the overall discrimination strength of the matching itemset with the ratio. For each target itemset in the scenario to be analyzed, obtain multiple matching itemset pairs corresponding to that target itemset. Determine the scenario discrimination coefficient of the target itemset based on the sum of the weighted discrimination coefficients of the multiple matching itemset pairs. For multiple target itemsets in the scenario to be analyzed, the target itemsets with a scenario discrimination coefficient greater than a preset threshold are determined as the preferred itemsets for the scenario to be analyzed, and the tags in the preferred itemsets are used as user profile features in the scenario to be analyzed.
6. A user profile construction system for apps based on multi-scenario fusion, characterized in that, The system includes: The analysis module is used to perform correlation analysis on sample data from multiple scenarios to obtain itemsets for each scenario; the sample data from multiple scenarios is obtained by dividing the merchant's transaction data into scenarios. The determination module is used to obtain the corresponding matching item set pairs based on the similarity between the target item set in the scene to be analyzed and the candidate item sets in other scenes, and to determine the common tags of the matching item set pairs and the unique tags of the target item set. The first calculation module is used to calculate the discrimination strength of the matching itemset pair based on the frequency of occurrence of the unique labels of the corresponding target itemset in different scenarios, determine the scenario discrimination strength of the matching itemset pair based on the difference in the frequency of occurrence of the common labels of the matching itemset pair, and determine the feature credibility of the matching itemset pair based on the proportion of sample data in the scenario corresponding to the matching itemset pair. The second calculation module is used to calculate the scene discrimination coefficient of each target itemset in the scene to be analyzed based on the discrimination strength, scene discrimination strength and feature credibility of the multiple matching itemset pairs corresponding to the target itemset, and to use the labels in the preferred itemset as user profile features in the scene to be analyzed; the preferred itemset indicates the itemset whose scene discrimination coefficient meets the preset conditions.
7. The APP user profile construction system based on multi-scenario fusion as described in claim 6, characterized in that, The analysis module is also used for: Based on preset scenario segmentation rules, the merchant's transaction data is classified into multiple data subsets; each data subset corresponds to a scenario. Association analysis algorithms are used to perform association analysis on sample data in each scenario to obtain at least one maximum frequent itemset in each scenario; where the itemset includes the maximum frequent itemset.
8. The APP user profile construction system based on multi-scenario fusion as described in claim 6, characterized in that, The determining module is also used for: Calculate the similarity between the target itemset in the scenario to be analyzed and multiple candidate itemsets in each other scenario, and select the candidate itemset with the highest similarity and greater than the preset filtering threshold. Associate the target itemset with the selected candidate itemset to obtain a matching itemset pair. The intersection of the two corresponding itemsets of a matching itemset pair is determined as the common label of the matching itemset pair; The unique labels of the target itemset are determined by the difference between the common labels of the target itemset and the matching itemset pair.
9. The APP user profile construction system based on multi-scenario fusion as described in claim 6, characterized in that, The first calculation module is also used for: Obtain the target itemset in the scenario to be analyzed, and obtain multiple matching itemset pairs formed by the target itemset and candidate itemsets in multiple other scenarios; For each matching item pair, obtain the unique label in the target item set, and count the frequency of occurrence of the unique label in the unique labels of multiple matching item pairs. Determine the distinguishing strength of the matching item pair based on the exponential decay value of the occurrence frequency. Obtain the common label of the matching itemset pair, calculate the first occurrence frequency of the common label in the sample data of the scene to be analyzed corresponding to the matching itemset pair, and the second occurrence frequency of the common label in the sample data of other scenes corresponding to the matching itemset pair, and determine the scene discrimination strength of the matching itemset pair based on the difference between the first occurrence frequency and the second occurrence frequency. The conversion rate of a scenario is determined by the ratio between the number of sample data in each scenario and the number of sample data across multiple scenarios. Determine the first scenario conversion rate of the matching itemset for the corresponding scenario to be analyzed, and determine the second scenario conversion rate of the matching itemset for the corresponding other scenarios. Calculate the first sum of the conversion rates of the first scenario and the conversion rates of the second scenario, and calculate the second sum of the first frequency of occurrence and the second frequency of occurrence; The feature confidence of the matching itemset pair is determined by the product of the first sum and the second sum.
10. The APP user profile construction system based on multi-scenario fusion as described in claim 6, characterized in that, The second calculation module is also used for: The overall discrimination strength of a matching itemset is obtained by multiplying the discrimination strength of each matching itemset pair by the scene discrimination strength of that matching itemset pair. The weighted discrimination coefficient of each matching itemset is obtained by calculating the ratio of the feature confidence of each matching itemset to the similarity of that matching itemset, and by multiplying the overall discrimination strength of the matching itemset with the ratio. For each target itemset in the scenario to be analyzed, obtain multiple matching itemset pairs corresponding to that target itemset. Determine the scenario discrimination coefficient of the target itemset based on the sum of the weighted discrimination coefficients of the multiple matching itemset pairs. For multiple target itemsets in the scenario to be analyzed, the target itemsets with a scenario discrimination coefficient greater than a preset threshold are determined as the preferred itemsets for the scenario to be analyzed, and the tags in the preferred itemsets are used as user profile features in the scenario to be analyzed.