Attribution analysis method and device based on multi-dimensional customer tag, equipment, and computer program product

Abstract

The application discloses a multi-dimensional customer label-based attribution analysis method and device, equipment and a computer program product. The method comprises the following steps: acquiring user data; performing multi-dimensional customer label classification processing and summarizing on the user data according to a time dimension to obtain processed user data; acquiring attribution analysis parameter information required by a business; calculating attribution indexes according to the processed user data and the attribution analysis parameter information to obtain a calculation result of the attribution indexes; and performing attribution analysis according to the calculation result of the attribution indexes to obtain an attribution analysis result. The application realizes dynamic correlation analysis on multi-dimensional customer attribute and behavior data. The attribution index calculation and analysis are combined with the processed user data and the attribution analysis parameter, so that key factors influencing business index changes can be located, and scientific and accurate decision support is provided for financial institutions in market promotion, customer service optimization and the like.

Description

Technical Field

[0001] This application relates to the field of financial data analysis technology, and in particular to an attribution analysis method, apparatus and equipment, and computer program product based on multidimensional customer tags. Background Technology

[0002] With the rapid development of fintech, the credit card business has entered a stage of digital and intelligent operation. Banks and financial institutions reach customers through multi-channel marketing (such as app push notifications, SMS, and offline activities), and optimize product design, adjust credit granting strategies, and improve user activity based on customer spending behavior data. In this process, attribution analysis technology has become a core tool for evaluating the effectiveness of various marketing channels and identifying key customer touchpoints, helping institutions quantify the contribution of different factors (such as advertising and incentives) to the final conversion (such as card opening, spending, and installment payments).

[0003] Currently, when analyzing the impact of various factors (such as payment channels, payment institutions, and customer credit limits) on final results (such as average daily spending, number of transactions, and number of active customers), financial institutions primarily rely on traditional attribution models (such as last-click attribution and linear attribution) or basic data-driven methods (such as Shapley values). However, these methods have significant limitations:

[0004] (1) Insufficient real-time performance: Most attribution analysis relies on offline batch processing, which makes it difficult to respond to changes in customer behavior in a timely manner, resulting in a lag in strategy adjustment.

[0005] (2) Disconnect between influencing factors and attribution results: As financial scenarios become more complex, the need to quantify cross-channel synergy is becoming increasingly prominent. For example, customer A has a credit limit of 100,000 yuan and habitually uses Alipay for consumption, with a monthly consumption amount of 80,000 yuan. Customer B has a credit limit of 10,000 yuan and habitually uses Tenpay for consumption, with a monthly consumption amount of 5,000 yuan. Traditional models may mistakenly attribute all the credit to the payment institution, Alipay, while underestimating the auxiliary role of credit limit size. This bias can lead to misallocation of marketing resources and even a decline in customer experience.

[0006] Therefore, how to achieve accurate, real-time, and interpretable attribution analysis while protecting user privacy has become a key technical challenge for improving the operational efficiency of credit card business. Summary of the Invention

[0007] This application provides an attribution analysis method, apparatus, and computer program product based on multidimensional customer tags to achieve accurate, real-time, and interpretable attribution analysis, providing financial institutions with dynamic and fine-grained decision support.

[0008] The embodiments of this application adopt the following technical solutions:

[0009] In a first aspect, embodiments of this application provide an attribution analysis method based on multidimensional customer tags, the attribution analysis method based on multidimensional customer tags including:

[0010] Acquire user data, including user consumption data and user activity data;

[0011] The user data is classified and summarized according to the time dimension using multi-dimensional customer tags to obtain processed user data, which includes user consumption data and user activity data.

[0012] Obtain the attribution analysis parameter information required for business operations;

[0013] Attribution indices are calculated based on the processed user data and the attribution analysis parameter information to obtain the calculation results of the attribution indices;

[0014] Attribution analysis is performed based on the calculation results of the attribution indices to obtain the attribution analysis results.

[0015] Optionally, obtaining user data includes:

[0016] The user data is obtained from the bank's core system;

[0017] The user data includes branch information, payment channels, payment institutions, consumption scenarios, new customer indicators, first transaction time, consumption city level, customer acquisition channels, credit limit, consumption amount, and active time.

[0018] Optionally, the user data is classified and summarized according to the time dimension using multi-dimensional customer tags to obtain processed user data, including:

[0019] The user data is preprocessed to obtain preprocessed user data, which includes standardized time.

[0020] Based on the preprocessed user data, multidimensional customer tags required for the business are generated according to the standardized time dimension and then summarized to obtain the summarized data.

[0021] The aggregated data is stored in a preset database.

[0022] Optionally, the attribution analysis parameter information required for obtaining the business includes:

[0023] Obtain attribution analysis parameter information of business configuration through the front-end interface;

[0024] The attribution analysis parameters include the analysis type, analysis indicators, and time range.

[0025] Optionally, the attribution analysis parameter information includes analysis type, analysis indicators, and time range. The analysis type includes activity type or consumption type, and the time range includes analysis time and comparison time. The calculation of attribution indicators based on the processed user consumption and activity data and the attribution analysis parameter information, to obtain the calculation results of the attribution indicators, includes:

[0026] Based on the attribution analysis parameter information, user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time are determined from the processed user data. The target customer group is either a customer group configured by the business or a default customer group when no configuration is made.

[0027] Based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time, the indicator values ​​of the analysis indicators and the measurement indicators are calculated as the calculation results of the attribution indicators.

[0028] Optionally, the measurement indicators include element surprise and element explanatory power. The step of calculating the indicator values ​​of the analysis indicators and the measurement indicators based on user data corresponding to the analysis indicators at the analysis time and the comparison time for the target customer group, and using these as the calculation results of the attribution indicators, includes:

[0029] Based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time, calculate the indicator values ​​of each dimension and element of the analysis indicators of the target customer group at the analysis time and the comparison time, as well as the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time.

[0030] Based on the index values ​​of the analysis indicators of the target customer group in each dimension and element at the analysis time and the comparison time, and the total index value of the analysis indicators of the target customer group at the analysis time and the comparison time, the index value of the element surprise power of the target customer group in each dimension and element is calculated using Js divergence.

[0031] Based on the index values ​​of the analytical indicators of the target customer group in each dimension and element at the analysis time and the comparison time, and the total index value of the analytical indicators of the target customer group at the analysis time and the comparison time, the index value of the explanatory power of the target customer group in each dimension and element is calculated.

[0032] Optionally, the measurement metrics further include dimensional surprise and dimensional explanatory power. The step of calculating the metric values ​​of the analysis metrics and the measurement metrics based on user data corresponding to the analysis metrics for the target customer group at the analysis time and the comparison time, and using these as the calculation results of the attribution metrics, includes:

[0033] Based on the index values ​​of the explanatory power of each element in each dimension of the target customer group, the elements are filtered using a preset element threshold to obtain the first filtered elements in each dimension.

[0034] Based on the element surprise power index value and element explanatory power index value of the first filtered elements in each dimension, the elements in each dimension are sorted according to the preset sorting rules to obtain the first element sorting results in each dimension.

[0035] The element surprise index values ​​of the first filtered elements in each dimension are summarized to obtain the dimension surprise index values ​​of the target customer group in each dimension.

[0036] Based on the first element sorting results of each dimension, the explanatory power index values ​​of the first-ranked elements in each dimension are summarized to obtain the explanatory power index values ​​of the target customer group in each dimension. Element filtering is performed using preset dimension thresholds to obtain the second filtered elements in each dimension.

[0037] Optionally, the attribution analysis based on the calculation results of the attribution index, to obtain the attribution analysis results, includes:

[0038] Based on the index values ​​of the dimensional surprise power of the target customer group in each dimension, the dimensions are sorted to obtain the dimensional sorting results for each dimension.

[0039] Based on the index value of the element explanatory power of the second filtered elements in each dimension, the elements in each dimension are sorted to obtain the second element sorting results for each dimension.

[0040] The root cause is determined based on the dimensional ranking results of each dimension.

[0041] The key influencing elements of the dimension in which the root cause resides are determined based on the sorting results of the second element of each dimension.

[0042] Secondly, embodiments of this application also provide an attribution analysis device based on multidimensional customer tags, the attribution analysis device based on multidimensional customer tags comprising:

[0043] The first acquisition unit is used to acquire user data, which includes user consumption data and user activity data.

[0044] The data processing unit is used to classify and summarize the user data according to the time dimension using multi-dimensional customer tags to obtain processed user data.

[0045] The second acquisition unit is used to acquire the attribution analysis parameter information required by the business.

[0046] The attribution calculation unit is used to calculate attribution indicators based on the processed user data and the attribution analysis parameter information, and obtain the calculation results of the attribution indicators.

[0047] The attribution analysis unit is used to perform attribution analysis based on the calculation results of the attribution indicators to obtain the attribution analysis results.

[0048] Thirdly, embodiments of this application also provide an apparatus, comprising:

[0049] A processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform any of the aforementioned attribution analysis methods based on multidimensional customer tags.

[0050] Fourthly, embodiments of this application also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements any of the aforementioned attribution analysis methods based on multidimensional customer tags.

[0051] The at least one technical solution adopted in this application embodiment can achieve the following beneficial effects: The attribution analysis method based on multidimensional customer tags in this application embodiment first acquires user data; then, the user data is classified and summarized according to the time dimension using multidimensional customer tags to obtain processed user data; then, the attribution analysis parameter information required by the business is acquired; then, attribution indicators are calculated based on the processed user data and attribution analysis parameter information to obtain the calculation results of the attribution indicators; finally, attribution analysis is performed based on the calculation results of the attribution indicators to obtain the attribution analysis results. This application embodiment realizes dynamic correlation analysis of multidimensional customer attributes and behavioral data by constructing a "time-tag" two-dimensional data model, solving the problem that traditional attribution models ignore the moderating role of customer characteristics. At the same time, it supports business personnel to flexibly configure attribution analysis parameters to meet the personalized needs of different business scenarios. By combining the processed user data and attribution analysis parameters for real-time calculation and analysis of attribution indicators, key factors affecting changes in business indicators can be identified, providing scientific and accurate decision support for banks and other financial institutions in market promotion, product design, and customer service optimization. Attached Figure Description

[0052] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0053] Figure 1 This is a flowchart illustrating an attribution analysis method based on multidimensional customer tags in an embodiment of this application.

[0054] Figure 2 This is a schematic diagram of the configuration interface for an attribution analysis parameter in an embodiment of this application;

[0055] Figure 3 This is a schematic diagram of the structure of an attribution analysis device based on multidimensional customer tags in an embodiment of this application;

[0056] Figure 4 This is a schematic diagram of the structure of a device according to an embodiment of this application. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0059] The technical terms used in this application mainly include:

[0060] (1) Attribution analysis: Attribution analysis is a methodology used to quantify the contribution of different influencing factors (such as payment channels, payment institutions, customer credit limits, etc.) to the final results (such as average daily consumption amount, number of transactions, number of active customers, etc.).

[0061] (2) Last click attribution: Attributing 100% of the conversion credit to the factors that the user finally touches. The method is simple but completely ignores the contribution of previous factors. It is suitable for short-term effect evaluation but not objective for long-term analysis.

[0062] (3) Linear attribution: It distributes the credit for conversion equally among all factors in the user path, acknowledging the value of multiple factors but failing to distinguish the actual differences in the influence of different factors.

[0063] (4) Data-driven attribution: Utilize machine learning or statistical models (such as Shapley values) to analyze user behavior data and dynamically quantify the contribution of each factor. This is suitable for complex scenarios but relies on big data support.

[0064] (5) Shapley value: a fair attribution method based on game theory, which allocates credit by calculating the marginal contribution of each factor in all possible combinations. The results are scientific but computationally complex.

[0065] (6) Js divergence (Jensen-Shannon Divergence): A method to measure the similarity between two probability distributions. It is used to calculate the element surprise. The smaller the value, the more similar the distributions are.

[0066] (7) Explanatory power of elements: measures the contribution of a specific element (such as "payment institution - Alipay") to the changes in the analysis indicators under a certain dimension. The calculation formula is the proportion of the difference in indicators of the element between the analysis date and the comparison date to the total difference in indicators.

[0067] (8) Element Surprise Power: Measures the distribution difference of a specific element under a certain dimension. It is obtained by calculating the difference in index distribution between the analysis date and the comparison date (based on Js divergence) and reflects the degree of abnormality of the element.

[0068] (9) Dimensional explanatory power: The sum of the explanatory power of the first N elements under a certain dimension, used to evaluate the overall contribution of that dimension to the changes in the analytical indicators.

[0069] (10) Dimensional Surprise Power: The sum of the elemental surprise power of all elements in a certain dimension, used to assess the degree of abnormality of that dimension.

[0070] (11) Target customer group: The customer group selected through specific conditions (such as credit limit, consumption behavior, etc.) for targeted attribution analysis.

[0071] (12) ClickHouse: A high-performance columnar database designed for real-time analysis. In attribution analysis, ClickHouse can achieve multi-dimensional aggregation of billions of data points in seconds, significantly improving analysis efficiency.

[0072] This application provides an attribution analysis method based on multidimensional customer tags, such as... Figure 1 The diagram shows a flowchart of an attribution analysis method based on multidimensional customer tags according to an embodiment of this application. The attribution analysis method based on multidimensional customer tags includes the following steps S110 to S150:

[0073] Step S110: Obtain user data, which includes user consumption data and user activity data.

[0074] User data can be obtained from the bank's core system, including fields such as branch information, payment channels, payment institutions, consumption scenarios, new customer indicators, first transaction time, consumption city level, customer acquisition channels, credit limit, consumption amount, and active time.

[0075] Step S120: The user data is classified and summarized according to the time dimension using multi-dimensional customer tags to obtain the processed user data.

[0076] Based on business needs, a series of multi-dimensional customer tags are predefined. For example, for branches, payment channels, payment institutions, consumption scenarios, new customer indicators, and customer acquisition channels, the data is directly categorized according to the corresponding fields in the raw data. For consumption amount stratification, user consumption amounts can be divided into corresponding levels according to different amount ranges set by the business. Credit limit stratification is similar, categorized according to preset limit ranges. Then, user data with the same tags is summarized and statistically analyzed according to the time dimension. For example, the total consumption amount through specific payment channels and in specific consumption scenarios within a certain time period, as well as the number of active customers, are statistically analyzed to obtain processed user consumption data and activity data, making the data more structured and analytically valuable.

[0077] Step S130: Obtain the attribution analysis parameter information required by the business.

[0078] Business users can flexibly configure attribution analysis parameters in the system according to their own business needs, such as analysis metrics and time ranges. This flexible configuration allows attribution analysis to accurately meet the needs of different business scenarios.

[0079] Step S140: Calculate the attribution index based on the processed user data and the attribution analysis parameter information to obtain the calculation result of the attribution index.

[0080] Based on the data processed in step S120 and the parameters configured in step S130, attribution indicators are calculated using specific calculation logic. For example, if the analysis indicator is consumption amount and the time range is set to day t1 and day t2, then by comparing the consumption amount data under different label dimensions (such as different branches, different consumption scenarios, etc.) on day t1 and day t2, the contribution of each label dimension to the change in consumption amount is calculated, and the calculation results of the attribution indicators are obtained.

[0081] Step S150: Perform attribution analysis based on the calculation results of the attribution index to obtain the attribution analysis results.

[0082] The attribution metrics calculated in step S140 are analyzed in depth. For example, by analyzing the contribution of each tag dimension to changes in spending, key factors leading to changes in spending are identified, such as whether a decline in spending in a specific consumption scenario is due to a decrease in new customers or a decrease in the frequency of purchases by existing customers. The final attribution analysis results provide a strong basis for business decisions.

[0083] This application's embodiments, by constructing a two-dimensional "time-tag" data model, achieve dynamic correlation analysis of multi-dimensional customer attributes and behavioral data, solving the problem that traditional attribution models neglect the moderating effect of customer characteristics. Simultaneously, it supports business personnel in flexibly configuring attribution analysis parameters to meet the personalized needs of different business scenarios. By combining processed user data and attribution analysis parameters for real-time calculation and analysis of attribution indicators, key factors influencing changes in business indicators can be identified, providing banks and other financial institutions with scientific and accurate decision support in areas such as market promotion, product design, and customer service optimization.

[0084] In some embodiments of this application, obtaining user data includes: obtaining the user data from the bank's core system; wherein, the user data includes branch information, payment channels, payment institutions, consumption scenarios, new customer indicators, first transaction time, consumption city level, customer acquisition channels, credit limits, consumption amounts, and active time.

[0085] User data can be obtained from the bank's core system, which serves as a crucial platform for storing and processing various types of business data and encompasses a wealth of user consumption-related information. When retrieving data from the bank's core system, specific fields can be extracted based on business analysis needs.

[0086] These fields include, but are not limited to, branch information, which identifies the specific bank branch where the user's purchase occurred, helping to analyze differences in business performance across branches in different regions; payment channel information, which distinguishes whether the user made the purchase through online channels (such as online banking and mobile banking) or offline channels (such as POS machine card swiping and counter payment), providing insight into user payment habits; payment institution information, which identifies whether the user used third-party payment platforms such as Alipay and WeChat Pay, or the bank's own payment tools, aiding in analyzing the competitive landscape of the payment market; consumption scenario information, covering different areas such as catering, shopping, tourism, and education, allowing for understanding of user consumption preferences and spending power in various scenarios; and new customer identification information, used to differentiate between new and existing customers. This data is useful for analyzing the effectiveness of new customer acquisition and the differences in spending behavior between new and existing customers; first-swipe time information, recording the time when a customer first uses their bank card for a transaction, is important for studying customer lifecycle and spending activity; city tier information, categorized according to the level of city development, allows analysis of user spending characteristics at different city tiers; customer acquisition channel information, such as online advertising, offline promotional activities, and customer referrals, helps evaluate the effectiveness of different customer acquisition channels; credit limit information, reflecting the credit limit granted to customers by the bank, allows analysis of the relationship between credit limit and spending behavior; spending amount information, directly reflecting the specific amount of each user's spending, is an important basis for analyzing spending scale and trends; and active time data reflects user activity.

[0087] In the actual acquisition process, the specific information dimensions to be acquired can be flexibly selected according to the specific business scenario and analysis purpose to meet diverse business analysis needs.

[0088] This application's embodiments acquire user data from the bank's core system, encompassing various key information fields, providing a comprehensive, accurate, and rich data foundation for subsequent attribution analysis. This data accurately and meticulously reflects users' consumption behavior and characteristics, making the attribution analysis based on this data more precise and in-depth. This helps banks and other financial institutions clearly understand the strengths and weaknesses in their business operations, enabling them to develop targeted market strategies, optimize product design, adjust resource allocation, improve customer satisfaction and loyalty, and enhance business efficiency and market competitiveness.

[0089] In some embodiments of this application, the step of classifying and summarizing the user data according to the time dimension using multidimensional customer tags to obtain processed user data includes: preprocessing the user data to obtain preprocessed user data, wherein the preprocessed user data includes standardized time; generating multidimensional customer tags required for the business according to the standardized time dimension based on the preprocessed user data and summarizing them to obtain summarized data; and storing the summarized data in a preset database.

[0090] The initial step involves preprocessing the raw user data, with core operations including date standardization. Time fields in the user data are uniformly converted to DATE type (e.g., YYYY-MM-DD format), and hour, minute, and second information is removed to ensure time granularity is accurate to the day. This step standardizes the time dimension of the data, avoiding analytical biases caused by differences in time formats.

[0091] Based on the preprocessed data, multi-dimensional customer tags required for business operations are generated according to the calendar day dimension. Some tags require hierarchical mapping processing. For example, the credit limit tag needs to be divided into ranges based on the customer's specific credit limit value, such as 0-5k, 5k-1w, 1w-2w, 2w-3w, 3w-5w, 5w+, etc. The same applies to the hierarchical classification of consumption amount. Other tags, such as branch, payment channel, and consumption scenario, are directly classified based on the original fields. Finally, a two-dimensional data model of "date-tag" is constructed, which supports the preliminary summary of analytical indicators (such as consumption amount, large consumption amount) of multi-dimensional combinations (such as "branch + payment institution + payment channel...") by date. For example, when summarizing consumption amount by date, all consumption records under the multi-dimensional combination of the day are accumulated; for large consumption amounts (≥5K), records that meet the conditions need to be filtered before summarizing.

[0092] The processed aggregated data is stored in a pre-defined database, such as Clickhouse, with the data date used as the partition key. This partitioning design means that when querying data within a specific date range, only the corresponding partition needs to be scanned, significantly reducing the amount of data scanned and improving the efficiency of aggregate queries.

[0093] This application's embodiments achieve efficient structured processing and storage of user data through standardized preprocessing of the time dimension, dynamic generation and hierarchical mapping of multi-dimensional labels, and database storage optimization based on partition keys. Date standardization ensures consistency in the time dimension, providing a reliable foundation for cross-date analysis; the two-dimensional label model supports flexible analysis with multi-dimensional combinations, accurately quantifying the impact of different label dimensions on business metrics; and the database's partitioned storage mechanism significantly improves aggregation query performance, enabling attribution analysis to quickly respond to business needs.

[0094] In some embodiments of this application, obtaining the attribution analysis parameter information required for the business includes: obtaining the attribution analysis parameter information configured for the business through a front-end interface; wherein, the attribution analysis parameter information includes analysis type, time range, and analysis indicators.

[0095] Obtaining the attribution analysis parameters required for business operations can be achieved through a front-end interface, which business personnel can configure flexibly. For example... Figure 2 The diagram shows a configuration interface for attribution analysis parameters in an embodiment of this application, which may include the following configurations:

[0096] (1) Analysis type configuration: The front-end interface provides options for analysis types such as "consumption" and "activity". Business personnel can select "consumption" if they want to analyze the influencing factors of consumption-related business, and select "activity" if they are concerned about factors related to user activity.

[0097] (2) Time range configuration: The time range setting includes two parts: analysis date and comparison date. The analysis date is used to set the target time point or time period to be analyzed, and the comparison date is used as the reference time point or time period for the baseline. Business personnel can specify these two dates separately through interactive components such as date picker on the front-end interface. For example, select 2025-04-11 as the analysis date and 2025-03-31 as the comparison date.

[0098] (3) Configuration of analysis indicators: Depending on the selected analysis type, the front-end interface will display the corresponding analysis indicator options. If the active analysis type is selected, analysis indicators such as the number of active customers will appear for business personnel to select; if the consumption analysis type is selected, analysis indicators such as consumption amount will be presented, and business personnel can select the required analysis indicators as needed.

[0099] (4) Target Customer Group Configuration: The front-end interface allows business personnel to select specific customer groups for analysis. For example, in consumer analysis, filtering settings can be configured for consumer customer groups, such as selecting branches that are not equal to Anhui Provincial Branch, channels that are not equal to UnionPay QuickPass and UnionPay QR Code, and new customer indicators that are equal to "yes" to define the target customer group. If business personnel do not perform the target customer group selection operation, the system will default to using the entire customer group as the analysis object.

[0100] This application embodiment allows business personnel to easily configure attribution analysis parameters through a front-end interface, greatly improving the flexibility and adaptability of attribution analysis. Business personnel can independently select analysis types, set time ranges, choose analysis indicators, and define target customer groups according to different business scenarios and needs, enabling attribution analysis to accurately meet diverse business analysis requirements. This configuration method not only simplifies the analysis process, reduces the intervention of technical personnel, and improves analysis efficiency, but also ensures that the analysis results are more targeted and practical, helping business personnel quickly and accurately identify key factors affecting changes in business indicators.

[0101] In some embodiments of this application, the attribution analysis parameter information includes analysis type, time range, and analysis indicators. The time range includes analysis time and comparison time. The step of calculating the attribution indicator based on the processed user data and the attribution analysis parameter information to obtain the calculation result of the attribution indicator includes: determining, based on the attribution analysis parameter information, user data corresponding to the analysis indicators of the target customer group at the analysis time and comparison time in the processed user data, wherein the target customer group is a customer group configured by the business or a default customer group in the absence of configuration; and calculating the indicator value of the analysis indicator and the indicator value of the measurement indicator based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and comparison time, as the calculation result of the attribution indicator.

[0102] When calculating attribution metrics, the process begins by precisely targeting specific user data based on parameters such as analysis type, analysis time (within the time range), comparison time, analysis metrics, and target customer group (or the default customer group, such as all customers, if not configured). Specifically, target customer groups can be selected from the processed data based on filtering criteria, such as specific branches, payment channels, or new customer indicators. Then, user consumption and activity data corresponding to the analysis metrics are extracted for each target customer group at the analysis and comparison times. For example, if the analysis metric is consumption amount, the consumption amount data for the target customer group during these two time periods is extracted.

[0103] After obtaining the consumption and activity data corresponding to the analytical indicators for the target customer group at the analysis and comparison times, the attribution indicators are calculated. On one hand, the indicator values ​​themselves are calculated. For activity-based analysis, indicators such as the number of active customers are calculated; for consumption-based analysis, indicators such as the amount spent are calculated. On the other hand, the values ​​of the measurement indicators are calculated, such as surprise and explanatory power. Surprise can be used to measure whether the changes in the analytical indicators of the target customer group compared to the comparison time exceed expectations; explanatory power helps to determine the degree to which each factor explains the changes in the analytical indicators. Through these calculations, the final attribution indicator results are obtained.

[0104] This application's embodiments achieve in-depth analysis of the reasons for changes in business indicators by accurately locating the corresponding data of analytical indicators for the target customer group within a specific time range and calculating the indicator values ​​of analytical and measurement indicators. It can accurately quantify the business performance of the target customer group at different times, as well as the degree of influence of various factors on changes in business indicators. This helps business personnel quickly identify key influencing factors, providing a scientific and quantitative basis for business decisions, and assisting enterprises in optimizing marketing strategies, rationally allocating resources, and improving business operational efficiency and effectiveness.

[0105] In some embodiments of this application, the measurement indicators include element surprise power and element explanatory power. The step of calculating the indicator values ​​of the analysis indicators and the measurement indicators based on user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time, and using these as the calculation results of the attribution indicators, includes: calculating the indicator values ​​of the analysis indicators for each dimension and element of the target customer group at the analysis time and the comparison time, and the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time, based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time; calculating the element surprise power indicator value of the target customer group in each dimension and element using Js divergence based on the indicator values ​​of the analysis indicators for each dimension and element of the target customer group at the analysis time and the comparison time, and the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time; and calculating the element explanatory power indicator value of the target customer group in each dimension and element based on the indicator values ​​of the analysis indicators for each dimension and element of the target customer group at the analysis time and the comparison time, and the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time.

[0106] For example, when the selected analysis indicator is consumption amount, the selected attribution analysis time is April 1, 2025 (t1), the comparison time is March 31, 2025 (t2), and the selected customer group is the entire customer group.

[0107] (1) Calculation of analytical indicators

[0108] Calculate the analytical indicators for the target customer group at time t1 for dimension i and element j respectively. The index value, the dimension i at time t2, and the analysis index of element j The indicator values, and the analysis indicators of the target customer group at time t1. Total index value and analysis index of time t2 The total index value.

[0109] The “dimension” here can be understood as the multi-dimensional customer “tags” mentioned in the previous embodiments, such as branch, payment institution, payment scenario, new customer marker, region, customer acquisition channel, consumption amount stratification, credit limit stratification, etc.

[0110] (2) Calculation of measurement indicators

[0111] 1) Elemental Amazing Power

[0112] Surprise factor represents the distributional difference between the analyzed value and the comparison value. The Js divergence value is introduced as the surprise factor value, and the calculation formula is as follows:

[0113] (1)

[0114] (2)

[0115] (3)

[0116] in, For example, on April 1, 2025, the dimension is the branch and the element is the consumption amount of the Beijing branch. This represents the total value of the analytical indicators for the target customer group on day t1, such as the total spending amount of the target customer group on April 1, 2025. For example, the analysis indicators for day t2 are dimension i and element j, such as the consumption amount of the Beijing branch as the dimension on March 31, 2025. This refers to the analytical metrics for the target customer group on day t2, such as the total spending amount of the target customer group on March 31, 2025. To analyze the proportion of date index values ​​to total index values ​​for a single element, This compares the proportion of a single element's date indicator value to the total indicator value. Let i be the surprise value of element j in dimension i.

[0117] 2) Explanatory power of elements

[0118] (4)

[0119] in, The explanatory power of an element in dimension i and element j, such as the explanatory power of an element in dimension i and element j in Beijing.

[0120] For example, the total consumption amount on April 1, 2025 is 1 billion, of which 200 million is consumed by the Beijing branch; the total consumption amount on March 31, 2025 is 1.2 billion, of which 300 million is consumed by the Beijing branch (both are simulated test data). To analyze the root cause of the decrease in consumption on April 1, 2025 compared to the previous day, we use the consumption amount on April 1, 2025 as the analysis indicator and the consumption amount on March 31, 2025 as the comparison indicator. Then:

[0121]

[0122]

[0123]

[0124] The explanatory power of the Beijing branch:

[0125] .

[0126] This application's embodiments comprehensively and deeply analyze the reasons for changes in business indicators of the target customer group at different times by accurately calculating and measuring indicators. The element surprise power can measure the degree of difference in the indicator distribution of each element between the analysis time and the comparison time, and the element explanatory power can reflect the degree of contribution of each element to the change of the overall indicator, providing a strong basis for subsequent attribution analysis.

[0127] In some embodiments of this application, the measurement metrics further include dimensional surprise power and dimensional explanatory power. The step of calculating the metric values ​​of the analysis metrics and the measurement metrics based on user data corresponding to the analysis metrics of the target customer group at the analysis time and the comparison time, as the calculation result of the attribution metrics, includes: filtering elements using a preset element threshold based on the element explanatory power metric values ​​of the target customer group for each element in each dimension to obtain first filtered elements for each dimension; sorting elements within each dimension according to a preset sorting rule based on the element surprise power metric values ​​and element explanatory power metric values ​​of the first filtered elements in each dimension to obtain first element sorting results for each dimension; summarizing the element surprise power metric values ​​of the first filtered elements in each dimension to obtain the dimensional surprise power metric value of the target customer group in each dimension; summarizing the explanatory power metric values ​​of the top-ranked elements within each dimension based on the first element sorting results to obtain the dimensional explanatory power metric value of the target customer group in each dimension; and filtering elements using a preset dimensional threshold to obtain second filtered elements for each dimension.

[0128] After calculating the element explanatory power and element surprise power based on the aforementioned embodiments, elements with explanatory power less than a predefined element threshold are first filtered out. Within the same dimension, elements are sorted in descending order of element surprise power first, then element explanatory power. Next, the dimension explanatory power and dimension surprise power are calculated. The dimension surprise power is the sum of the element surprise powers of all elements in that dimension, and the dimension explanatory power is the sum of the explanatory power of the top j elements within the same dimension. Within the same dimension, when the cumulative sum of element explanatory power (dimension explanatory power) reaches the dimension threshold, the remaining elements are filtered out. The filtered elements are then sorted in descending order of dimension surprise power, and within the same dimension, in descending order of element explanatory power. This entire process forms a closed loop of "pre-noise reduction → accurate sorting → accurate quantization → secondary screening → efficient positioning".

[0129] The specific implementation process is as follows:

[0130] (1) Filter out elements with “explanation power < element threshold” by using element threshold to obtain the first filtered elements of each dimension, thereby achieving noise reduction in advance and reducing invalid calculations and interference.

[0131] This step eliminates "meaningless elements" at the source, reducing redundancy in subsequent processes. Elements with explanatory power below the threshold are essentially minor objects with "minimal impact" on indicator fluctuations (e.g., a branch's consumption amount fluctuation accounts for only 1% of the dimension, below the element threshold of 5%). After pre-filtering, subsequent steps such as sorting, dimension indicator calculation, and dimension filtering only target elements with "actual impact," preventing low-value elements from consuming computing resources and interfering with analysis conclusions, thus significantly improving process execution efficiency.

[0132] On the other hand, if low-impact elements do not interfere with the ranking of high-value elements within the same dimension, some elements with low explanatory power but high surprise value (such as "a sudden fluctuation in traffic from a niche channel but with extremely low contribution") may occupy a high position in the ranking within the same dimension, obscuring the truly "influential and unexpected" core elements. After pre-filtering, the ranking objects are all elements that "reach the basic impact threshold", and the ranking results are more in line with the core goal of "finding the root cause".

[0133] On the other hand, it can reduce the risk of distortion in the calculation of dimensional indicators. The subsequent calculation of dimensional explanatory power (the sum of the explanatory power of the first j elements) and dimensional surprise power (the sum of the surprise power of the elements) is based on "effective elements" rather than "all elements", avoiding the dilution of the true influence of the dimension by low explanatory power elements (for example, if the core elements of a certain dimension originally contribute 80%, if a large number of low explanatory power elements are mixed in, the dimensional indicator will be lowered, leading to misjudgment).

[0134] (2) Within the same dimension, sort in descending order of “elements’ surprising power → elements’ explanatory power”. That is, sort the elements first according to the index value of elements’ surprising power, and then sort them according to the index value of elements’ explanatory power to obtain the sorting results of the first element in each dimension, thereby focusing on “high-value core elements” and avoiding subjective bias.

[0135] The core logic behind the above priority ranking rule for "impactful and unexpected" key elements is "surprise first, then impact":

[0136] High surprise factor = elemental changes exceeding expectations (such as a sudden plunge in prices from a previously stable channel), which is a "potential abnormal root cause";

[0137] High explanatory power = elements that contribute significantly to dimensional fluctuations (such as consumption amount accounting for 60% of a dimension) are "key influencing factors".

[0138] After sorting, elements that are "more unexpected and have a greater impact" are presented first within the same dimension, which solves the problem that "only looking at the degree of impact can easily miss sudden anomalies, and only looking at the unexpectedness can easily focus on ineffective fluctuations".

[0139] By replacing subjective judgment with quantitative rules, the objectivity of sorting is improved. There is no need to manually select "which element is more important". Through hierarchical sorting rules of "surprise + explanation", the priority of elements is automatically output, avoiding judgment bias caused by empiricism.

[0140] (3) Calculate the dimensional surprise and dimensional explanatory power.

[0141] 1) Dimensional Amazement

[0142] (5)

[0143] in, Let be the dimensional surprise of dimension i, such as the dimensional surprise of dimension i being a row, and n be the total number of elements after the first filter under dimension i.

[0144] The dimension of surprise power is quantified into the "overall degree of surprise". For example, the total surprise power of the "branch" dimension is 0.9, while that of the "consumption scenario" dimension is 0.3. It can be directly seen that the fluctuation of the "branch" dimension is more unexpected, so it is the priority for analysis.

[0145] 2) Dimensional Explanatory Power

[0146] Based on the ranking results of the first element in each dimension, the explanatory power of the dimension is calculated using the index values ​​of the explanatory power of the first j elements in each dimension:

[0147] (6)

[0148] in, This refers to the dimensional explanatory power of dimension i and element j, such as the dimensional explanatory power of a dimension where the dimension is a line break and the element is a Beijing line break. The elemental explanatory power of k elements in the same dimension.

[0149] Because the first j elements are the most critical elements within a dimension, the sum of the explanatory power index values ​​of the first j elements can accurately reflect the actual contribution of that dimension to the overall index, rather than an invalid accumulation of all elements. For example, if there are 10 elements in a dimension, but the sum of the explanatory power index values ​​of the first 2 elements accounts for 90% of the explanatory power, it means that the sum of the explanatory power of the first 2 elements can represent the core influence of that dimension.

[0150] (4) Use the dimension threshold to filter out elements with minimal influence in each dimension. That is, in a certain dimension, if the sum of the explanatory power of the current j elements (dimensional explanatory power) reaches the dimension threshold, the remaining elements can be filtered out.

[0151] On the one hand, it can eliminate analytical noise. Filtering out elements with minimal impact within a dimension (such as elements with less than 5% explanatory power) avoids wasting effort on objects that "have almost no effect on indicator fluctuations." On the other hand, it can narrow the scope of analysis, allowing subsequent focus only on elements that "truly explain indicator changes," thus improving the accuracy of conclusions.

[0152] In some embodiments of this application, the attribution analysis based on the calculation results of the attribution index to obtain the attribution analysis results includes: ranking the dimensions according to the index values ​​of the dimensional surprise power of the target customer group in each dimension to obtain the dimensional ranking results of each dimension; ranking the elements within each dimension according to the index values ​​of the element explanatory power of the second filtered elements in each dimension to obtain the second element ranking results of each dimension; determining the dimension where the root cause is located based on the dimensional ranking results of each dimension; and determining the key influencing elements of the dimension where the root cause is located based on the second element ranking results of each dimension.

[0153] Based on the calculation results of the aforementioned embodiments, elements are sorted in descending order of "dimensional surprise power → explanatory power of elements within the same dimension," constructing a clear root cause path and efficiently supporting decision-making. It should be noted that the sorting of explanatory power of elements within each dimension is based on the remaining elements after filtering using element and dimension thresholds in the aforementioned embodiments.

[0154] The final attribution analysis results can be presented in a "dimension-element" format, such as "branch - Beijing branch, Shanghai branch", "payment institution - JD.com", etc., clearly showing the contribution of specific elements under each dimension to the changes in indicators. At the same time, if the target customer group is limited, the analysis results will focus on the business characteristics of that specific group.

[0155] The above sorting rules are designed to make the global priorities clear and avoid disordered analysis. The first priority is "dimension surprise power in reverse order": prioritize the dimensions that "overall exceed expectations", such as the "consumption scenario" with sudden abnormalities. These dimensions often hide core problems / opportunities that have not been predicted and are the ones that the business needs to respond to most urgently. The second priority is "explanatory power of elements within the same dimension in reverse order": prioritize the elements with the "greatest impact" within the same dimension (such as "hotel travel" within "consumption scenario") to avoid disordered investigation within the dimension.

[0156] On the other hand, after achieving efficient tracing and sorting from "macro-dimensional to micro-element", a clear root cause path is formed, such as "high-surprise-capturing dimension: consumption scenario → high-explanatory-capturing element: hotel travel". Businesses do not need to analyze all dimensions and elements one by one. They can directly locate the core root cause based on the sorting results, which greatly improves decision-making efficiency.

[0157] On the other hand, the ranking results of supporting precise resource allocation directly guide resource allocation: prioritize the concentration of optimized resources on dimensions and elements with "high surprise value and high explanatory power", such as focusing on adjusting hotel and travel consumption promotion strategies, avoiding the dispersion of resources to low-value targets, and improving resource utilization efficiency.

[0158] In summary, the key points of this application are mainly as follows:

[0159] (1) This application designs a dynamic and scalable classification system of attribution influencing factors to achieve the goal of fine-grained interpretation of attribution analysis results.

[0160] This system covers multiple dimensions and levels in the credit card financial scenario (such as payment institutions, payment channels, credit limits, etc.), and is further refined into operable elements within each dimension (such as Tenpay, Alipay, Douyin, etc. under the payment institution dimension). Its innovation lies in the dynamic expansion mechanism, which supports adding new customer tag dimensions and specific elements under each dimension through configuration, without having to reconstruct the data model.

[0161] (2) This application designs a "date-tag" two-dimensional data model, which generates the tags required by the business according to the natural day dimension based on the preprocessed consumption flow, thereby achieving the purpose of cross-time attribution analysis.

[0162] Business personnel can freely select target and comparison dates. The system automatically aligns time granularity (by day / week / month / quarter / year) and allows for the free selection of specific analysis indicators. Then, by calculating attribution analysis metrics, the system quantifies the contribution of each label dimension and the elements under each dimension to the differences in analysis indicators, thereby identifying elements whose distribution deviates significantly from the historical benchmark.

[0163] (3) This application designs an attribution analysis calculation method based on element explanatory power, element surprising power, dimension explanatory power and dimension surprising power as measurement indicators, which solves the problems of dimension importance assessment and limited interpretation of results.

[0164] This application defines four dynamic indicators: element explanatory power, element surprising power, dimension explanatory power, and dimension surprising power. The influence of indicator changes is attributed to the dimension by ranking the magnitude of dimension surprising power, and the specific influencing element is attributed to the element by ranking the magnitude of element explanatory power, thereby obtaining the attribution analysis results quickly and accurately.

[0165] (4) This application designs an interactive and visual attribution report generation tool to support business personnel in dynamically exploring attribution results.

[0166] This application allows for real-time adjustment of information such as date range, target customer group, analysis indicators, and indicator thresholds. The system will automatically calculate the attribution analysis results in real time based on the above information, thereby obtaining attribution analysis results that drill down from the dimension level to the detailed element level.

[0167] The main technical effects of this application include:

[0168] (1) Fine-grained and dynamic: By constructing a two-dimensional data model of "date-tag" and a hierarchical classification system, dynamic correlation analysis of customer multidimensional attributes and behavioral data is realized, which solves the problem that traditional attribution models ignore the moderating effect of customer characteristics.

[0169] (2) Real-time performance and interpretability: Based on quantitative indicators such as element explanatory power and surprise power, combined with the efficient aggregation capabilities of the Clickhouse database, it supports real-time calculation of multi-dimensional contributions and outputs intuitive "dimensional-element" root cause combinations. Compared with the traditional offline batch processing method, it significantly reduces the lag in strategy adjustment.

[0170] (3) Standardization and scalability: From data collection and tag mapping to attribution calculation, the entire process follows preset rules to ensure consistent analysis logic across different scenarios and times. It supports the rapid addition of analysis indicators, dimensions, or elements through configuration without modifying the underlying data model or code, ensuring flexible adaptation to future needs.

[0171] This application also provides an attribution analysis device 300 based on multidimensional customer tags, such as... Figure 3 The diagram shows a structural schematic of an attribution analysis device based on multidimensional customer tags according to an embodiment of this application. The attribution analysis device 300 based on multidimensional customer tags includes: a first acquisition unit 310, a data processing unit 320, a second acquisition unit 330, an attribution calculation unit 340, and an attribution analysis unit 350, wherein:

[0172] The first acquisition unit 310 is used to acquire user data, which includes user consumption data and user activity data.

[0173] The data processing unit 320 is used to classify and summarize the user data according to the time dimension using multi-dimensional customer tags to obtain the processed user data.

[0174] The second acquisition unit 330 is used to acquire the attribution analysis parameter information required by the business.

[0175] The attribution calculation unit 340 is used to calculate attribution indicators based on the processed user data and the attribution analysis parameter information, and obtain the calculation results of the attribution indicators.

[0176] The attribution analysis unit 350 is used to perform attribution analysis based on the calculation results of the attribution indicators to obtain attribution analysis results.

[0177] In some embodiments of this application, the first acquisition unit 310 is specifically used to: acquire the user data from the bank's core system; wherein, the user data includes branch information, payment channels, payment institutions, consumption scenarios, new customer indicators, first transaction time, consumption city level, customer acquisition channels, credit limits, consumption amounts, and active time.

[0178] In some embodiments of this application, the data processing unit 320 is specifically used for: preprocessing the user data to obtain preprocessed user data, the preprocessed user data including standardized time; generating multidimensional customer tags required for business according to the dimension of standardized time based on the preprocessed user data and performing summary processing to obtain summarized data; and storing the summarized data in a preset database.

[0179] In some embodiments of this application, the second acquisition unit 330 is specifically used to: acquire attribution analysis parameter information of business configuration through a front-end interface; wherein, the attribution analysis parameter information includes analysis type, time range and analysis indicators.

[0180] In some embodiments of this application, the attribution analysis parameter information includes analysis type, time range, and analysis indicators. The time range includes analysis time and comparison time. The attribution calculation unit 340 is specifically used to: determine, based on the attribution analysis parameter information, user data corresponding to the analysis indicators of the target customer group at the analysis time and comparison time from the processed user data, wherein the target customer group is a customer group configured by the business or a default customer group in the absence of configuration; and calculate the indicator values ​​of the analysis indicators and the measurement indicators based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and comparison time, as the calculation results of the attribution indicators.

[0181] In some embodiments of this application, the measurement indicators include element surprise power and element explanatory power. The attribution calculation unit 340 is specifically used to: calculate the indicator values ​​of the analysis indicators of the target customer group for each dimension and element at the analysis time and the comparison time, respectively, based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time; calculate the indicator value of the element surprise power of the target customer group in each dimension and element at the analysis time and the comparison time using Js divergence, based on the indicator values ​​of the analysis indicators of the target customer group for each dimension and element at the analysis time and the comparison time, and the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time; and calculate the indicator value of the element explanatory power of the target customer group in each dimension and element at the analysis time and the comparison time, based on the indicator values ​​of the analysis indicators of the target customer group for each dimension and element at the analysis time and the comparison time, and the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time.

[0182] In some embodiments of this application, the measurement indicators further include dimensional surprise power and dimensional explanatory power. The attribution calculation unit 340 is specifically used for: filtering elements based on the explanatory power index values ​​of each element in each dimension of the target customer group using a preset element threshold to obtain the first filtered elements in each dimension; sorting the elements in each dimension according to a preset sorting rule based on the dimensional surprise power index values ​​and the explanatory power index values ​​of the first filtered elements in each dimension to obtain the first element sorting result in each dimension; summarizing the dimensional surprise power index values ​​of the first filtered elements in each dimension to obtain the dimensional surprise power index value of the target customer group in each dimension; summarizing the explanatory power index values ​​of the first elements in each dimension based on the first element sorting result in each dimension to obtain the dimensional explanatory power index value of the target customer group in each dimension; and filtering elements using a preset dimensional threshold to obtain the second filtered elements in each dimension.

[0183] In some embodiments of this application, the attribution analysis unit 350 is specifically used for: sorting dimensions according to the index values ​​of the dimensional surprise power of the target customer group in each dimension, and obtaining the dimensional sorting results of each dimension; sorting elements within each dimension according to the index values ​​of the element explanatory power of the second filtered elements in each dimension, and obtaining the second element sorting results of each dimension; determining the dimension where the root cause is located according to the dimensional sorting results of each dimension; and determining the key influencing elements of the dimension where the root cause is located according to the second element sorting results of each dimension.

[0184] It is understood that the above-described attribution analysis device based on multidimensional customer tags can realize each step of the attribution analysis method based on multidimensional customer tags provided in the foregoing embodiments. The relevant explanations of the attribution analysis method based on multidimensional customer tags are applicable to the attribution analysis device based on multidimensional customer tags, and will not be repeated here.

[0185] Figure 4 This is a schematic diagram of the structure of a device according to an embodiment of this application. For example... Figure 4 As shown, the device includes one or more processors (or processing units), and may also include one or more memories coupled to the processors, and may also include a communication module coupled to the processors.

[0186] A communication module can be used to communicate with other devices or apparatuses, such as sending or receiving data and / or signals. A communication module may have at least one communication module for communication. A communication module may include any interface necessary for communicating with other devices. Exemplarily, a communication module may be a transceiver, circuit, bus, module, or other type of communication module.

[0187] The processor may include, but is not limited to, one or more of the following: a general-purpose computer, a special-purpose computer, a microcontroller, a digital signal processor (DSP), or a controller-based multi-core controller architecture. The device may have multiple processors, such as application-specific integrated circuit (ASIC) chips, which are time-dependent on a clock synchronized with the main processor.

[0188] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), electrically programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD), or other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that do not persist during the duration of a power outage.

[0189] A computer program consists of computer-executable instructions that are executed by an associated processor. Programs can be stored in ROM. A processor can perform any appropriate action and processing by loading the program into RAM.

[0190] Possible implementations of this application can be achieved through a program, enabling the communication device to execute any of the processes discussed in the foregoing embodiments. Possible implementations of this application can also be achieved through hardware or a combination of software and hardware.

[0191] In some implementations, the program may be tangibly contained in a computer-readable storage medium, which may include in a device (such as in memory) or other storage device accessible by the device. The program may be loaded from the computer-readable storage medium into RAM for execution. The computer-readable storage medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, CD, DVD, etc.

[0192] This application also provides a computer-readable storage medium storing computer instructions or program code thereon, which, when executed by a processor, causes the processor to perform the methods and functions involved in any of the above embodiments. A computer-readable medium can be any tangible medium that contains or stores a program for or relating to an instruction execution system, apparatus, or device. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. More detailed examples of computer-readable storage media include electrical connections with one or more wires, magnetic media (e.g., disks, floppy disks, hard disks, magnetic tapes, magnetic storage devices), optical media (e.g., optical storage devices, DVDs), semiconductor media (e.g., solid-state drives), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), or any suitable combination thereof.

[0193] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. Embodiments of this application also provide at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. This computer program product includes one or more computer-executable instructions, such as instructions included in a program module, which execute in a device on a target real or virtual processor to perform the processes, methods, and functions involved in any of the above embodiments. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0194] This application also proposes a computer program product, including a computer program or instructions that, when run on a computer, cause the computer to perform the processes, methods, and functions described in the above embodiments. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided as needed. The machine-executable instructions for the program modules can be executed locally or in a distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0195] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software, which can be executed by a controller, microprocessor, or other computing device. Although various aspects of the embodiments of this disclosure are shown and described as block diagrams, flowcharts, or represented using some other illustration, it should be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as, as non-limiting examples, in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0196] It should be noted that although embodiments of this application have been described above with reference to the accompanying drawings, these embodiments are not independent of each other, and they can be combined to obtain other embodiments. The methods, situations, categories, and classifications of embodiments in this application are only for the convenience of description and should not constitute a special limitation. Various methods, categories, situations, and features in embodiments can be combined with each other if logically consistent. The various embodiments of this application can be arbitrarily combined to achieve different technical effects. The embodiments of this application will not list various combinations.

[0197] Furthermore, although the operation of the methods of this disclosure is described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowcharts may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps. It should also be noted that the features and functions of two or more devices according to this disclosure may be embodied in one device. Conversely, the features and functions of one device described above may be further divided and embodied by multiple devices.

[0198] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0199] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for attribution analysis based on multi-dimensional customer tags, characterized in that, The attribution analysis method based on multidimensional customer tags includes: Acquire user data, including user consumption data and user activity data; The user data is classified and summarized according to the time dimension using multi-dimensional customer tags to obtain the processed user data; Obtain the attribution analysis parameter information required for business operations; Attribution indices are calculated based on the processed user data and the attribution analysis parameter information to obtain the calculation results of the attribution indices; Attribution analysis is performed based on the calculation results of the attribution indices to obtain the attribution analysis results. 2.The method of claim 1, wherein, The acquisition of user data includes: The user data is obtained from the bank's core system; The user data includes branch information, payment channels, payment institutions, consumption scenarios, new customer indicators, first transaction time, consumption city level, customer acquisition channels, credit limit, consumption amount, and active time. 3.The method of claim 1, wherein, The process of classifying and summarizing the user data according to the time dimension using multi-dimensional customer tags results in the following processed user data: The user data is preprocessed to obtain preprocessed user data, which includes standardized time. Based on the preprocessed user data, multidimensional customer tags required for the business are generated according to the standardized time dimension and then summarized to obtain the summarized data. The aggregated data is stored in a preset database. 4.The method of claim 1, wherein, The attribution analysis parameter information required for obtaining the business includes: Obtain attribution analysis parameter information of business configuration through the front-end interface; The attribution analysis parameters include the analysis type, time range, and analysis indicators. 5.The method of claim 1, wherein, The attribution analysis parameter information includes analysis type, time range, and analysis indicators. The analysis type includes active or consumer categories, and the time range includes analysis time and comparison time. The calculation of attribution indicators based on the processed user data and the attribution analysis parameter information yields the following results: Based on the attribution analysis parameter information, user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time are determined from the processed user data. The target customer group is either a customer group configured by the business or a default customer group when no configuration is made. Based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time, the indicator values ​​of the analysis indicators and the measurement indicators are calculated as the calculation results of the attribution indicators. 6.The method of claim 5, wherein, The measurement metrics include element surprise and element explanatory power. The calculation of the analysis metric values ​​and the measurement metric values ​​based on user data corresponding to the analysis metrics for the target customer group at the analysis time and the comparison time, as the calculation results of the attribution metrics, includes: Based on the user data corresponding to the analysis indicators of the target customer group at the analysis time and the comparison time, calculate the indicator values ​​of each dimension and element of the analysis indicators of the target customer group at the analysis time and the comparison time, as well as the total indicator value of the analysis indicators of the target customer group at the analysis time and the comparison time. Based on the index values ​​of the analysis indicators of the target customer group in each dimension and element at the analysis time and the comparison time, and the total index value of the analysis indicators of the target customer group at the analysis time and the comparison time, the index value of the element surprise power of the target customer group in each dimension and element is calculated using Js divergence. Based on the index values ​​of the analytical indicators of the target customer group in each dimension and element at the analysis time and the comparison time, and the total index value of the analytical indicators of the target customer group at the analysis time and the comparison time, the index value of the explanatory power of the target customer group in each dimension and element is calculated. 7.The method of claim 6, wherein, The measurement metrics also include dimensional surprise and dimensional explanatory power. The calculation of the analysis metric values ​​and the measurement metric values ​​based on user data corresponding to the analysis metrics for the target customer group at the analysis time and the comparison time, as the calculation results of the attribution metrics, includes: Based on the index values ​​of the explanatory power of each element in each dimension of the target customer group, the elements are filtered using a preset element threshold to obtain the first filtered elements in each dimension. Based on the element surprise power index value and element explanatory power index value of the first filtered elements in each dimension, the elements in each dimension are sorted according to the preset sorting rules to obtain the first element sorting results in each dimension. The element surprise index values ​​of the first filtered elements in each dimension are summarized to obtain the dimension surprise index values ​​of the target customer group in each dimension. Based on the first element sorting results of each dimension, the explanatory power index values ​​of the first-ranked elements in each dimension are summarized to obtain the explanatory power index values ​​of the target customer group in each dimension. Element filtering is performed using preset dimension thresholds to obtain the second filtered elements in each dimension.

8. The attribution analysis method based on multidimensional customer tags according to claim 7, characterized in that, The attribution analysis performed based on the calculation results of the attribution indices, to obtain the attribution analysis results, includes: Based on the index values ​​of the dimensional surprise power of the target customer group in each dimension, the dimensions are sorted to obtain the dimensional sorting results for each dimension. Based on the index value of the element explanatory power of the second filtered elements in each dimension, the elements in each dimension are sorted to obtain the second element sorting results for each dimension. The root cause is determined based on the dimensional ranking results of each dimension. The key influencing elements of the dimension in which the root cause resides are determined based on the sorting results of the second element of each dimension.

9. An attribution analysis device based on multidimensional customer tags, characterized in that, The attribution analysis device based on multidimensional customer tags includes: The first acquisition unit is used to acquire user data, which includes user consumption data and user activity data. The data processing unit is used to classify and summarize the user data according to the time dimension using multi-dimensional customer tags to obtain processed user data. The second acquisition unit is used to acquire the attribution analysis parameter information required by the business. The attribution calculation unit is used to calculate attribution indicators based on the processed user data and the attribution analysis parameter information, and obtain the calculation results of the attribution indicators. The attribution analysis unit is used to perform attribution analysis based on the calculation results of the attribution indicators to obtain the attribution analysis results.

10. An apparatus comprising: processor; And a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the attribution analysis method based on multidimensional customer tags as described in any one of claims 1 to 8.

11. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the attribution analysis method based on multidimensional customer tags as described in any one of claims 1 to 8.

Images