User consumption behavior monitoring and analysis method and system based on network summary computing

Abstract

The present invention belongs to the field of big data technology, and is a user consumption behavior monitoring and analysis method and system based on the summary calculation of the network connection, comprising the following steps: S1: obtaining the historical consumption behavior data of the bank network connection, and obtaining the keywords in the consumption behavior data; S2: establishing a consumption theme system based on the keywords, and deriving consumption behavior indicators. S3: obtaining the user's bank network connection consumption behavior data, extracting keywords, and matching the corresponding consumption themes; S4: obtaining the corresponding consumption behavior indicators, calculating the quantitative values ​​of each consumption behavior indicator and storing them; S5: extracting the historical quantitative values ​​of the same consumption behavior indicator of the user and comparing them with the new quantitative values, judging whether the user's consumption behavior is abnormal based on the change amplitude and time interval of the quantitative values, and monitoring abnormal users. It solves the problem that the channels for small and medium-sized banks to obtain user consumption behavior in the prior art are narrow and the content obtained is limited, resulting in the inability to monitor and analyze user consumption behavior.

Description

Technical Field

[0001] The present invention belongs to the field of big data technology, and specifically relates to a user consumption behavior monitoring and analysis method and system based on network summary calculation. Background Art

[0002] With the rapid development of emerging technology industries such as artificial intelligence, cloud computing, and big data, more and more industries are using these emerging technologies to classify users and analyze user needs in order to provide precise services to users. The financial industry is no exception. More and more banks are classifying customers according to customer needs and paying attention to the refined development of marketing products. Among them, how to use user account history data, user browsing history and other information about consumer behavior to further strengthen the effective identification and classification of user needs and user characteristics has become an important issue in the banking industry.

[0003] The user behavior monitoring and analysis of large banks depends on their mobile banking APP tracking points and APP user activity. However, for some small and medium-sized banks, due to the poor activity of their mobile banking APPs, the complex content of mobile banking, limited functions, and limited tracking content and technology for capturing behavior, it is impossible to accurately collect and analyze user consumption behavior data through factors such as mobile banking APP tracking points and user activity, which limits small and medium-sized banks' real-time monitoring and mining analysis of customer behavior. Therefore, a new user consumption behavior monitoring and analysis method is urgently needed to solve the problem that small and medium-sized banks cannot accurately and comprehensively monitor and analyze user consumption behavior under existing technologies. Summary of the invention

[0004] The present invention aims to provide a user consumption behavior monitoring and analysis method and system based on NetsUnion summary calculation, which solves the problem in the prior art that small and medium-sized banks have narrow channels for obtaining user consumption behavior and limited content, resulting in the inability to accurately and comprehensively monitor and analyze user consumption behavior.

[0005] The basic solution provided by the present invention: a user consumption behavior monitoring and analysis method based on network summary calculation, comprising the following steps:

[0006] S1: Obtain the historical consumption behavior data of the bank's online banking and obtain the keywords in the consumption behavior data;

[0007] S2: Establish a consumption theme system based on keywords, wherein the consumption theme system includes primary consumption tags, secondary consumption themes, and tertiary consumption themes; combine the tertiary consumption themes with consumption behavior data, restore consumption behavior scenarios, and generate consumption behavior indicators;

[0008] S3: Obtain the user's bank network consumption behavior data, obtain keywords in the consumption behavior data, and match the corresponding third-level consumption topics according to the keywords;

[0009] S4: Obtain the consumption behavior indicators under the third-level consumption theme, and calculate the quantitative values ​​of each consumption behavior indicator based on the consumption behavior data; save each consumption behavior indicator, quantitative value and transaction time in the user consumption behavior record;

[0010] S5: extract the historical quantitative value of the same consumption behavior indicator of the user and compare it with the new quantitative value of the consumption behavior indicator, and judge whether the consumption behavior of the user is abnormal according to the change range and time interval of the quantitative value. If so, monitor the user.

[0011] The principles and advantages of the present invention are as follows: the present invention proposes a user consumption behavior monitoring and analysis method based on network summary calculation, which uses network consumption behavior data to mine user consumption keywords and establishes a complete consumption theme system based on the keywords; by tracking the consumption behavior data of a specified user, the user's third-level consumption theme is judged, and the user's consumption behavior data and consumption behavior indicators are combined to obtain a quantitative value of the consumption behavior indicator, and by comparing the historical quantitative value with the new quantitative value and the transaction time, it is analyzed and judged whether the user's consumption behavior is abnormal, and abnormal users are monitored.

[0012] The beneficial effects of this solution are: it utilizes the activeness of online consumption behavior to obtain user consumption behavior data, alleviating the problem of small and medium-sized banks having narrow channels for obtaining consumer behavior and limited content; and by constructing a consumption theme system and the concept of quantifying consumer behavior indicators, it realizes real-time identification and analysis of user transaction behavior; it compares the user's specific consumption behavior with his or her historical consumption behavior to identify abnormalities in the user's consumption behavior, solving the problem of small and medium-sized banks being unable to monitor and analyze user consumption behavior.

[0013] Furthermore, the step S2 further comprises the following steps:

[0014] S2-1: Combine keywords to form several primary consumption tags;

[0015] S2-2: Divide the primary consumption labels into different consumption themes, including secondary consumption themes and tertiary consumption themes;

[0016] S2-3: Combine the three-level consumption themes with consumption behavior data, restore consumption behavior scenarios, form consumption behavior indicators, and extract the indicator feature vector and feature judgment basis of each consumption behavior indicator;

[0017] S2-4: Put the indicator feature vector and specific target behavior of the consumer behavior indicator into the analysis model for analysis, construct the core behavior pattern, the sample characteristics and abnormal standards of each core behavior pattern, and at the same time generate the associated behavior of the core behavior pattern and the associated behavior judgment logic. The associated behavior judgment logic is used to determine whether the consumer behavior indicator is an associated behavior under the core behavior pattern.

[0018] Beneficial effects: Keywords are first combined to form several primary consumption tags, and then the primary consumption tags are divided into different consumption themes to form a consumption theme system, thereby improving the closeness and relevance of keywords under each consumption theme, and improving the accuracy of consumption themes matched according to keywords.

[0019] Further, the step S3 specifically includes the following steps:

[0020] S3-1: Obtain the user's bank network consumption behavior data, obtain the keywords in the transaction summary, and count the frequency of each keyword in the summary and the corresponding consumption theme;

[0021] S3-2: Determine consumption themes based on the extracted keywords, convert keywords into word vectors, and convert each consumption theme into a theme feature vector;

[0022] S3-3: Based on the frequency of occurrence of keywords, the word vector is similarly calculated with the feature vector of each topic, and the degree of consistency between the consumption behavior and each consumption topic is calculated based on the similarity;

[0023] S3-4: Determine the consumption theme with the highest degree of consistency as the consumption theme of the consumption behavior; determine the core behavior pattern of the consumption behavior based on the consumption behavior indicator characteristics under the consumption theme, and determine whether the consumption behavior is an associated behavior under the core behavior pattern based on the associated behavior judgment logic.

[0024] Beneficial effects: Keywords and consumption themes are converted into vector values, and the degree of consistency between consumption behavior and each consumption theme is calculated based on the frequency of occurrence of keywords. The consumption theme with the highest degree of consistency is selected to improve the accuracy of consumption theme matching.

[0025] Further, the step S4 specifically includes the following steps:

[0026] S4-1: Determine the user group based on the consumption behavior indicator, and calculate the quantitative value of the consumption behavior indicator based on the consumption behavior data;

[0027] S4-2: Save the consumption behavior indicators, quantitative values ​​and consumption behavior data to the user's consumption behavior record, and update the sample characteristics of the core behavior pattern and the distribution characteristics of the user group;

[0028] S4-3: Obtain all consumption behavior indicators in the user's consumption behavior, and obtain the distribution of the user's consumption behavior indicators and the distribution of the group's consumption behavior indicators;

[0029] S4-4: Determine the core behavior pattern based on the distribution of consumption behavior indicators of users and groups and the sample characteristics of the core behavior pattern, indicator feature vectors, related behaviors, and distribution characteristics of the user group; the core behavior pattern is divided into conventional, investment, and qualification categories; the related behaviors include but are not limited to financial management, product purchases, user defaults, multiple investment and lending risks, and user ratings.

[0030] Beneficial effects: Introducing the concept of core behavior patterns and classifying users according to their core behavior patterns will help small and medium-sized banks to conduct subsequent management, analysis and monitoring of users.

[0031] Furthermore, the step S5 further comprises the following steps:

[0032] S5-1: extract the historical quantitative value of the same consumption behavior indicator of the user, the group distribution characteristics of the third-level consumption theme corresponding to the consumption behavior indicator, and the core behavior pattern related behavior judgment logic, and compare the historical quantitative data of the consumption behavior indicator, the group distribution characteristics of the third-level consumption theme corresponding to the consumption behavior indicator, and the core behavior pattern related behavior judgment logic with the new quantitative value; obtain the change range and time interval of the quantitative value;

[0033] S5-2: Obtain the core behavior pattern of the user. If the core behavior pattern of the user is a regular type, execute step S5-3; if the core behavior pattern of the user is an investment type, execute step S5-4; if the core behavior pattern of the user is a qualification type, execute step S5-5;

[0034] S5-3: In the same time interval, determine whether the change range of the user's quantitative value exceeds the set range. If so, mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal;

[0035] S5-4: within the same time interval, determine whether the change range of the user's quantitative value reaches a set value. If so, mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal;

[0036] S5-5: Obtain the quantitative value of the same consumption behavior indicator of other users with the same core behavior pattern as the user. In the same time interval, if the change range of the quantitative value of the consumption behavior indicator of the user exceeds or is lower than N% of the quantitative value of the same consumption behavior indicator of the user's group in the time interval, then mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal;

[0037] S5-6: Monitor the marked users and push user information according to the trigger rules.

[0038] Beneficial effects: According to different core behavior patterns, the user's consumption behavior abnormality judgment standard is adjusted to improve the accuracy of the abnormal judgment results. When the user's core behavior pattern is qualification type, the user's consumption behavior is abnormally identified through a longitudinal comparison of the user's historical consumption behavior and a horizontal comparison with the consumption behavior of the user's group, thereby improving the flexibility of abnormal judgment. Compared with the existing technology, the abnormal judgment results of this solution are more accurate and more in line with the user's actual situation.

[0039] Furthermore, the triggering rules are as follows:

[0040] If the user's core behavior pattern is a regular one, record the user's third-level consumption tag subject category and quantitative value, and pass the aggregated user list and third-level consumption tags to the relevant system for reference;

[0041] If the user's core behavior pattern is investment, obtain the user information and store it for marketing business use;

[0042] If the user's core behavior pattern is qualification-based, the user-related data will be transmitted to the corresponding decision engine for corresponding rule verification.

[0043] Beneficial effects: Different trigger rules are triggered according to the core behavior patterns of users, so that small and medium-sized banks can analyze and monitor customers based on different needs, manage and serve different customers, and provide customers with accurate and beneficial help.

[0044] The present invention also provides a user consumption behavior monitoring and analysis system based on network summary calculation, including a topic establishment subsystem, a behavior quantification subsystem and an anomaly identification and monitoring subsystem;

[0045] The topic establishment subsystem includes a data acquisition module, a keyword extraction module, a corpus system and quantification module, and a feature extraction and vector family derivation module;

[0046] The data acquisition module is used to acquire the consumption behavior data of the bank's network connection history, and the consumption behavior data includes transaction summary, transaction time, transaction amount and transaction frequency;

[0047] The keyword extraction module is used for keywords in consumer behavior data and classifies the keywords to form a number of primary consumer tags;

[0048] The corpus system and quantification module are used to convert the three-level consumption tags into word vectors under several consumption themes, build the relationship between the consumption themes between the word vectors, and generate a consumption theme system;

[0049] The feature extraction and vector family derivation module is used to combine each consumption theme with the subordinate consumption behavior data, derive a number of consumption behavior indicators, and construct the indicator feature vector and feature determination basis of the consumption behavior indicator;

[0050] The behavior quantification subsystem includes a collection module, a matching module, a calculation module and a storage module;

[0051] The collection module is used to collect the user's bank network consumption behavior data and obtain keywords in the transaction summary;

[0052] The matching module is used to match the consumption theme corresponding to the consumption behavior according to the keywords;

[0053] The calculation module is used to obtain the consumption behavior indicators under the consumption theme, and calculate the quantitative values ​​of each consumption behavior indicator according to the consumption behavior data;

[0054] The storage module is used to store various consumption behavior indicators, quantitative values ​​and transaction time of the user;

[0055] The abnormality identification and monitoring subsystem includes a comparison module, a judgment module, a marking module and a monitoring module;

[0056] The comparison module is used to extract the historical quantitative value of the same consumption behavior indicator of the user and compare it with the new quantitative value of the consumption behavior indicator to obtain the change range and time interval of the quantitative value;

[0057] The judgment module is used to judge whether the consumption behavior indicator of the user is abnormal according to the change range and time interval of the quantitative value;

[0058] The marking module is used to mark users with abnormal consumption behavior;

[0059] The monitoring module is used to monitor the marked users.

[0060] The technical principles and advantages of this solution are: keyword recognition based on the historical consumption behavior data of the network, and the establishment of a consumption theme system based on keywords, the user's consumption theme is calculated based on the probability of occurrence of keywords in the user's consumption behavior data, and the quantitative values ​​of the user's consumption behavior indicators are calculated and saved based on the user's consumption behavior data, and the user's consumption behavior is analyzed for abnormalities in combination with the quantitative values ​​of the user's historical consumption behavior indicators, and users with abnormal consumption behavior are monitored. This solution alleviates the difficulties of small and medium-sized banks in obtaining users' consumption behavior through narrow channels and limited content, and solves the problem that small and medium-sized banks cannot monitor and analyze user consumption behavior.

[0061] Furthermore, the quantification subsystem also includes an indicator distribution module and a classification module;

[0062] The quantification subsystem also includes an indicator distribution module and a classification module;

[0063] The indicator distribution module is used to obtain all consumption behavior indicators in the user's consumption behavior and obtain the distribution of the user's consumption behavior indicators;

[0064] The classification module is used to identify the user's core behavior pattern based on the user's consumption behavior indicator distribution and the user's associated behavior; the core behavior pattern is divided into conventional, investment and qualification categories; the associated behaviors include financial management, product purchase, user default and user rating.

[0065] Beneficial effects: Introducing the concept of core behavior patterns and classifying users according to their core behavior patterns will help small and medium-sized banks to conduct subsequent management, analysis and monitoring of users.

[0066] Furthermore, the anomaly identification and monitoring subsystem also includes a pattern acquisition module and a rule recording module;

[0067] The pattern acquisition module is used to acquire the core behavior pattern of the user; the rule recording module is used to record the abnormal judgment rules and trigger rules;

[0068] The abnormality judgment rules are as follows: General category: if the change range of the user's quantitative value exceeds the set range, it is marked as abnormal; Investment category: if the change range of the user's quantitative value reaches the set value, it is marked as abnormal; Qualification category: if the change range of the quantitative value of the user's consumption behavior indicator exceeds or is lower than N% of the quantitative value of the consumption behavior indicator of the group of users with the same core behavior pattern, it is marked as abnormal;

[0069] The trigger rules are as follows: General category: record the consumption tag of the user, and pass the aggregated user list and consumption tag to the relevant system for reference; Investment category: obtain user information for storage for marketing business use; Qualification category: store user-related data and transmit it to the corresponding decision engine for corresponding rule verification;

[0070] The judgment module is also used to judge whether the consumption behavior indicator of the user is abnormal according to the abnormal judgment rules and the core behavior pattern of the user;

[0071] The monitoring module is used to monitor the marked users according to the core behavior patterns and triggering rules of the users.

[0072] Beneficial effects: According to different core behavior patterns, the user's consumption behavior abnormality judgment standard is adjusted to improve the accuracy of the abnormal judgment results. When the user's core behavior pattern is qualification type, the user's consumption behavior is abnormally identified through a longitudinal comparison of the user's historical consumption behavior and a horizontal comparison with the consumption behavior of the user's group, thereby improving the flexibility of abnormal judgment. Compared with the existing technology, the abnormal judgment results of this solution are more accurate and more in line with the user's actual situation. At the same time, different trigger rules are triggered according to the user's core behavior pattern, so that small and medium-sized banks can analyze and monitor customers based on different needs, which is beneficial for small and medium-sized banks to manage and serve different customers, so that customers can get accurate and beneficial help. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a structural schematic diagram of an embodiment of a user consumption behavior monitoring and analysis system based on network summary calculation according to the present invention.

[0074] Figure 2 Schematic diagram of the steps of an embodiment of the method for monitoring and analyzing user consumption behavior based on network summary calculation of the present invention. DETAILED DESCRIPTION

[0075] The following is further described in detail through specific implementation methods:

[0076] The implementation example is as follows: A user consumption behavior monitoring and analysis method based on network summary calculation uses the following Figure 1 The user consumption behavior monitoring and analysis system based on network summary calculation shown in the figure includes a theme establishment subsystem, a behavior quantification subsystem and an anomaly identification and monitoring subsystem;

[0077] The topic establishment subsystem includes a data acquisition module, a keyword extraction module, a corpus system and quantification module, and a feature extraction and vector family derivation module;

[0078] The data acquisition module is used to acquire the consumption behavior data of the bank's network connection history, and the consumption behavior data includes transaction summary, transaction time, transaction amount and transaction frequency;

[0079] The keyword extraction module is used for keywords in consumer behavior data and classifies the keywords to form a number of primary consumer tags;

[0080] The corpus system and quantification module are used to divide the primary consumption tags into word vectors of several consumption themes, construct the relationship between the consumption themes between the word vectors, and generate a consumption theme system; in this embodiment, all transaction summaries are segmented, cut, stop words are removed, frequent words are searched, descriptive word conversion, high-frequency synonym mapping, TF-IDF word frequency statistics and part-of-speech combination are performed through Chinese word segmentation technology to extract a total of 217 keywords, and the keywords are classified to form 95 consumption tags and form 37 consumption themes.

[0081] The feature extraction and vector family derivation module is used to combine each consumption theme with the affiliated consumption behavior data, restore the consumption behavior scenario (including theme, scenario, amount, frequency, time, user identity, category, channel, etc.), generate consumption behavior indicators, and construct the indicator feature vector and feature judgment basis of the consumption behavior indicator; in this embodiment, 950 consumption behavior indicator systems are derived based on consumption themes combined with consumption behavior data such as transaction time, transaction frequency, and transaction amount.

[0082] The behavior quantification subsystem includes a collection module, a matching module, a calculation module, a storage module, an indicator distribution module and a classification module;

[0083] The collection module is used to collect the user's bank network consumption behavior data and obtain keywords in the transaction summary;

[0084] The matching module is used to match the consumption theme corresponding to the consumption behavior according to the keywords; in this embodiment, the specific matching process is as follows: based on the set of consumption themes of the UnionPay transaction, a regular expression is constructed with 217 keywords, phrase recognition is performed on the summary field of the user's UnionPay transaction record, the hit of each phrase is counted and converted into the corresponding keyword, and the degree of fit between the consumption behavior and the consumption theme is calculated according to the hit frequency and hit threshold of the keywords and the similarity of the theme feature vector, and the consumption theme with the highest fit and the minimum fit standard is returned as the consumption theme of the transaction.

[0085] The calculation module is used to obtain the consumption behavior indicators under the consumption theme, and calculate the quantitative values ​​of each consumption behavior indicator according to the consumption behavior data;

[0086] The storage module is used to store various consumption behavior indicators, quantitative values ​​and transaction time of the user;

[0087] The indicator distribution module is used to obtain all consumption behavior indicators in the user's consumption behavior and obtain the distribution of the user's consumption behavior indicators;

[0088] The classification module is used to identify the core behavior patterns of users according to the distribution of consumption behavior indicators of users and the associated behaviors of users; the core behavior patterns are divided into conventional, investment and qualification categories; the associated behaviors include but are not limited to financial management, product purchases, user defaults, user ratings, multiple loans and capital needs; the classification module also uses association rule analysis, structured equation analysis, time series, distributed lag models, etc. to identify the consumption formation patterns of users;

[0089] The anomaly identification and monitoring subsystem includes a comparison module, a judgment module, a marking module, a monitoring module, a pattern acquisition module and a rule recording module;

[0090] The comparison module is used to extract the historical quantitative value of the same consumption behavior indicator of the user, the group distribution characteristic value of the user group and compare it with the new quantitative value of the consumption behavior indicator to obtain the change range and time interval of the quantitative value;

[0091] The pattern acquisition module is used to acquire the core behavior pattern of the user;

[0092] The rule recording module is used to record abnormality judgment rules and trigger rules;

[0093] The abnormality judgment rules are as follows: General category: if the change range of the user's quantitative value exceeds the set range, it is marked as abnormal; Investment category: if the change range of the user's quantitative value reaches the set value, and the probability calculated by the corresponding investment behavior probability model reaches the threshold standard, it is marked as abnormal; Qualification category: if the change range of the quantitative value of the user's consumption behavior indicator exceeds or is lower than N% of the quantitative value of the consumption behavior indicator of the group of users with the same core behavior pattern, it is marked as abnormal; in this embodiment, the quantitative value of the consumption behavior indicator of the group to which the user belongs is the percentile value of the quantitative group distribution, where the value range of N is 5 to 20;

[0094] The trigger rules are as follows: General category: record the consumption tag of the user, and pass the aggregated user list and consumption tag to the relevant system for reference; Investment category: obtain user information for storage, and call the corresponding marketing purchase behavior probability calculation model. If the model calculation standard is passed, the result will be transmitted to the downstream business system for marketing business use; Qualification category: store the user-related data in the corresponding risk module according to the segmentation type, and transmit it to the corresponding decision engine for corresponding rule verification;

[0095] The judgment module is also used to judge whether the consumption behavior indicator of the user is abnormal according to the abnormal judgment rules and the core behavior pattern of the user;

[0096] The marking module is used to mark users with abnormal consumption behavior;

[0097] The monitoring module is used to monitor the marked users according to the core behavior patterns and triggering rules of the users.

[0098] As attached Figure 2 As shown: A user consumption behavior monitoring and analysis method based on network summary calculation includes the following steps:

[0099] S1: Obtain the historical consumption behavior data of the bank's online banking service, obtain the phrases in the consumption behavior data, and convert them into keywords;

[0100] S2-1: Combine keywords to form several primary consumption tags;

[0101] S2-2: Divide the primary consumption tags into different consumption themes, including secondary consumption themes and tertiary consumption themes, and construct the subject word data and subject feature vectors of consumption themes at different levels;

[0102] S2-3: Combine the three-level consumption themes with consumption behavior data, restore consumption behavior scenarios, form consumption behavior indicators, and extract the indicator feature vector and feature judgment basis of each consumption behavior indicator;

[0103] S2-4: Put the indicator feature vector and specific target behavior of the consumer behavior indicator into the analysis model for analysis, construct the core behavior pattern, the sample characteristics of each core behavior pattern and the abnormal standard, and generate the associated behavior of the core behavior pattern and the associated behavior judgment logic, the associated behavior judgment logic is used to determine whether the consumer behavior indicator is an associated behavior under the core behavior pattern; the analysis model in this embodiment includes a correlation analysis model, an association rule model, a structured equation model and a distributed lag model;

[0104] S3-1: Obtain the user's bank online payment consumption behavior data, obtain the phrases in the transaction summary, convert them into keywords, and count the frequency of each keyword in the summary and the corresponding consumption theme;

[0105] S3-2: Determine consumption themes based on the extracted keywords, convert keywords into word vectors and construct theme feature vectors, and convert each consumption theme into a theme feature vector;

[0106] S3-3: Based on the frequency of occurrence of the keywords, the word vector is similarly calculated with the feature vector of each topic, and the degree of consistency between the consumption behavior and each consumption topic is calculated based on the similarity; in this embodiment, the calculation method includes the cosine similarity of the third-level topic and the second-level topic classification algorithm;

[0107] S3-4: Determine the consumption theme with the highest degree of fit and that meets the minimum fit standard as the consumption theme of the consumption behavior; the minimum fit standard is to exclude the situation where "several candidate themes are very low and do not belong to any theme concept, but a theme with the highest relative value is forcibly derived"; determine the core behavior pattern of the consumption behavior based on the consumption behavior indicator characteristic judgment basis under the consumption theme, and determine whether the consumption behavior is an associated behavior under the core behavior pattern based on the associated behavior judgment logic;

[0108] S4-1: Determine the user group based on the consumption behavior indicator, and calculate the quantitative value of the consumption behavior indicator based on the consumption behavior data;

[0109] S4-2: Save the consumption behavior indicators, quantitative values ​​and consumption behavior data to the user's consumption behavior record, and update the sample characteristics of the core behavior pattern and the distribution characteristics of the user group;

[0110] S4-3: Obtain all consumption behavior indicators in the user's consumption behavior, and obtain the distribution of the user's consumption behavior indicators and the distribution of the group's consumption behavior indicators;

[0111] S4-4: Determine the core behavior pattern based on the distribution of consumption behavior indicators of users and groups, sample characteristics of the core behavior pattern, indicator feature vector, associated behavior, and distribution characteristics of the group to which the user belongs; the core behavior pattern is divided into conventional, investment, and qualification categories; the associated behaviors include but are not limited to financial management, product purchase, user default, multiple investment and lending risks, and user ratings;

[0112] S5-1: extract the historical quantitative value of the same consumption behavior indicator of the user, the group distribution characteristics of the third-level consumption theme corresponding to the consumption behavior indicator, and the core behavior pattern related behavior judgment logic, and compare the historical quantitative data of the consumption behavior indicator, the group distribution characteristics of the third-level consumption theme corresponding to the consumption behavior indicator, and the core behavior pattern related behavior judgment logic with the new quantitative value; obtain the change range and time interval of the quantitative value;

[0113] S5-2: Obtain the core behavior pattern of the user. If the core behavior pattern of the user is a regular type, execute step S5-3; if the core behavior pattern of the user is an investment type, execute step S5-4; if the core behavior pattern of the user is a qualification type, execute step S5-5;

[0114] S5-3: In the same time interval, determine whether the change range of the user's quantitative value exceeds the set range. If so, mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal;

[0115] S5-4: within the same time interval, determine whether the change range of the user's quantitative value reaches a set value. If so, mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal;

[0116] S5-5: Obtain the quantitative value of the same consumption behavior indicator of other users with the same core behavior pattern as the user. In the same time interval, if the change range of the quantitative value of the consumption behavior indicator of the user exceeds or is lower than 20% of the quantitative value of the same consumption behavior indicator of the user's group in the time interval, then mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal;

[0117] S5-6: Monitor the marked users and push user information according to the trigger rules.

[0118] The specific implementation process is as follows:

[0119] For example, the transaction flow of customer Zhang San's online payment on August 8, 2021 is as follows:

[0120] Transaction 1: Summary "Payment channel 1 consumption - City A Rail Transit Co., Ltd.", amount: X1, time: T1;

[0121] Transaction 2: Summary "Payment channel 1 consumption-XX Fund Management Co., Ltd.", amount: X2, time: T2;

[0122] Transaction 3: Summary "Payment channel 2 consumption - Supermarket C, City B", amount: X3, time: T3;

[0123] Transaction 4: Summary "Payment channel 3 consumption - D City Y Microfinance Co., Ltd.", amount: X4, time: T4;

[0124] Transaction 5: Summary "Payment channel 2 consumption - E online car-hailing company", amount: X5, time: T5;

[0125] After word segmentation, word segmentation, stop word processing, etc., the keywords of each transaction flow are:

[0126] Transaction 1: Payment channel 1, consumption, City A, rail, transportation, Co., Ltd.

[0127] Transaction 2: Payment channel 1, consumption, XX, fund, management, limited company;

[0128] Transaction 3: Payment channel 2, consumption, City B, Supermarket C;

[0129] Transaction 4: Payment channel 3, consumption, D city, small loan, limited company;

[0130] Flow 5: Payment channel 2, consumption, E, online car-hailing, travel;

[0131] After semantic conversion (synonym conversion, ordinal word expansion) and label recognition, the labels and values ​​of each transaction flow are as follows:

[0132] Flow 1: City A, Rail, Travel, Public Transportation, Value: X1;

[0133] Flow 2: Fund, Financial Management, Value: X2;

[0134] Flow 3: Takeaway, catering, consumption, value: X3;

[0135] Flow 4: City B, Y, micro-loan, credit, value: X4;

[0136] Flow 5: E, travel, taxi, value: X5;

[0137] After word vector conversion, the word vectors of each transaction flow are:

[0138] Flow 1: (0, 0, 0.05, 0.7, 0.6, 0, 0, ...);

[0139] Flow 2: (0, 0.99, 0, 0.1, 0.7, 0.6, 0, 0, ...);

[0140] Flow 3: (0.32, 0, 0.85, 0, 0, 0.63, 0, 0, …);

[0141] Flow 4: (0, 0.02, 0, 0, 0, 0.04, 0, …);

[0142] Flow 5: (0, 0, 0, 0.3, 0.99, 0.04, 0, 0, ...);

[0143] The word vectors of each flow are classified with the secondary theme feature vectors, and the similarity is calculated with the tertiary theme feature vectors. The consumption themes of each flow are calculated as follows:

[0144] Stream 1: Consumption theme: behavioral characteristics / transportation / public transportation-railway;

[0145] Flow 2: Consumption theme: product transactions / investment and financial management transactions / product purchase behavior-funds;

[0146] Flow 3: Consumption theme: catering consumption / dining out / take-out;

[0147] Transaction 4: Consumption theme: product transaction / loan transaction / small loan purchase;

[0148] Flow 5: Consumption theme: Behavioral characteristics / Transportation / Transportation-Taxi;

[0149] The transaction flow is judged against the consumption behavior indicator rules under the consumption theme to which it belongs, and the consumption behavior indicator is supplemented and improved in combination with the customer's historical consumption behavior data and the set basic data:

[0150] Transaction 1: Public transportation - rail: number of urban transportation transactions in the past month, indoor transportation transaction amount in the past month, month-on-month comparison of common transportation transactions in the past month, number of commuting transactions on weekdays in the past month, number of commuting transactions during non-rush hour on weekdays in the past month, maximum transaction amount of urban transportation in the past three months, average number of transportation transactions in the past three months...;

[0151] Transaction 2: Product purchase behavior - Funds: total fund purchase amount in the past three months, number of financial management purchases in the past three months, fund purchase channel preference in the past three months, number of Internet financial management in the past six months, types of fund products purchased in the past 12 months, types of financial management products purchased in the past 12 months...;

[0152] Transaction 3: Takeaway: number of takeaway transactions in the past month, takeaway transaction amount in the past month, number of takeaway transactions in the past three months, maximum takeaway transaction amount in the past three months, average number of takeaway transactions in the past six months, takeaway merchant preference in the past six months...;

[0153] Transaction 4: Small loan purchase: small loan transaction amount in the past month, number of small loan transactions in the past three months, type of small loan products, number of small loans held in the past 12 months, number of months from the last small loan, whether there are multiple small loans in the past three months, whether there are multiple small loans in the past three months, preference for online loan channels...;

[0154] Transaction 5: Transportation - Taxi: Number of taxi transactions in the past month, taxi transaction finance in the past month, taxi platform preference, month-on-month taxi growth in the past month, average number of taxi transactions in the past three months, average taxi transaction amount in the past three months, maximum taxi transaction amount in the past three months, maximum number of taxi transactions in the past three months, number of taxi rides on weekdays in the past three months, number of taxi rides on holidays in the past three months, number of taxi rides during peak hours on weekdays in the past three months...;

[0155] Comprehensively analyze the current / accumulative changes in consumer behavior indicators caused by each transaction flow based on customer historical behavior data and group characteristics distribution, determine whether the monitoring indicators of the core target behavior are met, and whether to trigger the corresponding monitoring behavior:

[0156] Flow 1: Among the indicator data updated after this transaction, the relevant core target behaviors include "product holding / auto installment loan". The main performance indicators of the target behavior and the indicator triggering rules of the related indicators are calculated based on the customer's current flow behavior and historical behavior data, as well as the distribution of the behavior in the customer's group. The following are: a significant increase in travel demand (including a significant increase in the average number of traffic trips in the past three months, a significant month-on-month increase in commuting during peak working hours in the past three months, a significant increase in the number of commuting during non-peak working hours in the past month, strong travel demand, and the percentile of traffic trips in the group...). The indicator triggering rule monitoring result score of "auto installment loan" is used to determine whether the core target behavior of "auto installment loan" is triggered.

[0157] Flow 2: After the fund for the product purchase behavior in which the flow is located updates the relevant indicator data, the relevant target behaviors include "product holding / investment and financial management purchase". The main performance indicators of the target behavior and the indicator triggering rules of the related indicators are calculated based on the customer's current flow behavior and historical behavior data, as well as the distribution of the behavior in the customer's group. The following are: a significant increase in investment and financial management demand (including the total amount of fund purchases in the past three months, the number of financial management purchases in the past three months, financial management purchase channel preferences, fund purchases in the past month, Internet financial management records in the past month, the interval between the last financial management transactions, financial management product type preferences...). According to the monitoring result score of the indicator triggering rule of "investment and financial management purchase" (the score calculation basis: the correlation of each indicator with the financial management product purchase behavior and the structured equation contribution index), it is determined whether the core target behavior of "financial management product purchase" is triggered.

[0158] Flow 3: Small amount of daily consumption, which does not cause changes in core target behavior, and only requires updating basic indicators;

[0159] Flow 4: The process of purchasing micro loans is similar to the above. The related target behaviors are "user capital demand" and "poor risk control - multiple loans". Based on the customer's current flow behavior and historical behavior data, the distribution of behavior in the customer's group, etc., the main performance indicators of the target behavior and the indicator triggering rules of the related indicators are calculated to determine whether the core target behavior is triggered.

[0160] Flow 5: See Flow 1;

[0161] Calculate the abnormal range of core consumer behaviors, perform abnormal judgment, and execute subsequent push behaviors according to the trigger rules:

[0162] Flow 1: If the core target behavior of "auto installment loan" is triggered, the analysis module of the target behavior is activated, including the "auto installment demand" calculation model and personalized auto installment product marketing. A high score triggers the model to be executed in combination with other customer basic data (the model has fixed characteristic indicator variables, and the latest characteristic indicator variables are brought into the model to calculate the willingness to pay for auto installments. If the willingness to pay for auto installments reaches the threshold, it is output to the auto installment marketing list to update the relevant data table to the downstream marketing system and trigger subsequent marketing activities). Personalized auto installment product marketing is similar to this, calculating the products, interest rates and installment plans that are suitable for users. This model is not within the scope of this patent and will not be presented.

[0163] Flow 2: If the core target behavior of "financial product purchase" is triggered, the analysis module of the target behavior will be activated, including the "investment and financial management marketing" and "fund cross-selling" measurement models to calculate the user's purchase intention and product preference within the industry. If the intention is greater than the threshold, it will enter the marketing list, update the relevant data table to the downstream marketing system, and trigger subsequent marketing activities.

[0164] Transaction 3: Small amount of daily living consumption is a regular behavior. You only need to update the indicators and label information under the scope of the third-level theme.

[0165] Flow 4: If "user capital demand" or "poor risk control - multiple borrowings" is triggered, the analysis module of the target behavior will be activated, which is "loan marketing" and "credit risk index". Combined with other data, the corresponding willingness or probability of poor risk control is calculated, and the corresponding marketing list or risk control blacklist is output.

[0166] Flow 5: See Flow 1.

[0167] The above are only embodiments of the present invention. Common knowledge such as the known specific structures and characteristics in the scheme is not described in detail here. Ordinary technicians in the relevant field know all the common technical knowledge in the technical field to which the invention belongs before the application date or priority date, can obtain all the existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the scope of protection of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A user consumption behavior monitoring and analysis method based on network summary calculation, characterized by: The following steps are involved: S1: Obtain the historical consumption behavior data of the bank's online banking and obtain the keywords in the consumption behavior data; S2: Establish a consumption theme system based on keywords, wherein the consumption theme system includes primary consumption tags, secondary consumption themes, and tertiary consumption themes; combine the tertiary consumption themes with consumption behavior data, restore consumption behavior scenarios, and generate consumption behavior indicators; S3: Obtain the user's bank network consumption behavior data, obtain keywords in the consumption behavior data, and match the corresponding third-level consumption topics according to the keywords; S4: Obtain the consumption behavior indicators under the third-level consumption theme, and calculate the quantitative values ​​of each consumption behavior indicator based on the consumption behavior data; save each consumption behavior indicator, quantitative value and transaction time in the user consumption behavior record; S5: extracting the historical quantitative value of the same consumption behavior indicator of the user and comparing it with the new quantitative value of the consumption behavior indicator, judging whether the consumption behavior of the user is abnormal according to the change range and time interval of the quantitative value, and if so, monitoring the user; Wherein, the step S2 includes the following sub-steps: S2-1: Combine keywords to form several primary consumption tags; S2-2: Divide the primary consumption labels into different consumption themes, including secondary consumption themes and tertiary consumption themes; S2-3: Combine the three-level consumption themes with consumption behavior data, restore the consumption behavior scenarios, form consumption behavior indicators, and extract the indicator feature vector and feature judgment basis of each consumption behavior indicator; S2-4: Put the indicator feature vector and specific target behavior of the consumer behavior indicator into the analysis model for analysis, construct the core behavior pattern, the sample characteristics of each core behavior pattern and the abnormal standard, and generate the associated behavior of the core behavior pattern and the associated behavior determination logic, wherein the associated behavior determination logic is used to determine whether the consumer behavior indicator is an associated behavior under the core behavior pattern; The step S3 comprises the following sub-steps: S3-1: Obtain the user's bank network consumption behavior data, obtain the keywords in the transaction summary, and count the frequency of each keyword in the summary and the corresponding consumption theme; S3-2: Determine consumption themes based on the extracted keywords, convert keywords into word vectors, and convert each consumption theme into a theme feature vector; S3-3: Based on the frequency of occurrence of keywords, the word vector is similarly calculated with the feature vector of each topic, and the degree of consistency between the consumption behavior and each consumption topic is calculated based on the similarity; S3-4: Determine the consumption theme with the highest degree of fit as the consumption theme of the consumption behavior; determine the core behavior pattern of the consumption behavior based on the consumption behavior indicator characteristics under the consumption theme, and determine whether the consumption behavior is an associated behavior under the core behavior pattern based on the associated behavior judgment logic; The step S4 comprises the following sub-steps: S4-1: Determine the user group based on the consumption behavior indicator, and calculate the quantitative value of the consumption behavior indicator based on the consumption behavior data; S4-2: Save the consumption behavior indicators, quantitative values ​​and consumption behavior data to the user's consumption behavior record, and update the sample characteristics of the core behavior pattern and the distribution characteristics of the user group; S4-3: Obtain all consumption behavior indicators in the user's consumption behavior, and obtain the distribution of the user's consumption behavior indicators and the distribution of the group's consumption behavior indicators; S4-4: Determine the core behavior pattern based on the distribution of consumption behavior indicators of users and groups, sample characteristics of the core behavior pattern, indicator feature vector, associated behaviors, and distribution characteristics of the user group; the core behavior pattern is divided into conventional, investment, and qualification categories; the associated behaviors include financial management, product purchase, user default, multiple investment and lending risks, and user ratings; The step S5 comprises the following sub-steps: S5-1: extract the historical quantitative value of the same consumption behavior indicator of the user, the group distribution characteristics of the third-level consumption theme corresponding to the consumption behavior indicator, and the core behavior pattern related behavior judgment logic, and compare the historical quantitative data of the consumption behavior indicator, the group distribution characteristics of the third-level consumption theme corresponding to the consumption behavior indicator, and the core behavior pattern related behavior judgment logic with the new quantitative value; obtain the change range and time interval of the quantitative value; S5-2: Obtain the core behavior pattern of the user. If the core behavior pattern of the user is a regular type, execute step S5-3; if the core behavior pattern of the user is an investment type, execute step S5-4; if the core behavior pattern of the user is a qualification type, execute step S5-5; S5-3: In the same time interval, determine whether the change range of the user's quantitative value exceeds the set range. If so, mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal; S5-4: within the same time interval, determine whether the change range of the user's quantitative value reaches a set value. If so, mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal; S5-5: Obtain the quantitative value of the same consumption behavior indicator of other users with the same core behavior pattern as the user. In the same time interval, if the change range of the quantitative value of the consumption behavior indicator of the user exceeds or is lower than N% of the quantitative value of the same consumption behavior indicator of the user's group in the time interval, then mark the user's consumption behavior as abnormal and execute step S5-6; otherwise, do not mark it as abnormal; S5-6: Monitor the marked users and push user information according to the trigger rules.

2. The user consumption behavior monitoring and analysis method based on network summary calculation according to claim 1 is characterized by: The triggering rules are as follows: If the user's core behavior pattern is a regular one, record the user's third-level consumption tag subject category and quantitative value, and pass the aggregated user list and third-level consumption tags to the relevant system for reference; If the user's core behavior pattern is investment, obtain the user information and store it for marketing business use; If the user's core behavior pattern is qualification-based, the user-related data will be transmitted to the corresponding decision engine for corresponding rule verification.

3. A user consumption behavior monitoring and analysis system based on network summary calculation, characterized by: It includes the subject establishment subsystem, behavior quantification subsystem and anomaly identification and monitoring subsystem; The topic establishment subsystem includes a data acquisition module, a keyword extraction module, a corpus system and quantification module, and a feature extraction and vector family derivation module; The data acquisition module is used to acquire the consumption behavior data of the bank's network connection history, and the consumption behavior data includes transaction summary, transaction time, transaction amount and transaction frequency; The keyword extraction module is used for keywords in consumer behavior data and classifies the keywords to form a number of primary consumer tags; The corpus system and quantification module are used to convert the primary consumption labels into word vectors under several consumption themes, construct the relationship between the consumption themes between the word vectors, and generate a consumption theme system; The feature extraction and vector family derivation module is used to combine each consumption theme with the subordinate consumption behavior data, derive a number of consumption behavior indicators, and construct the indicator feature vector and feature determination basis of the consumption behavior indicator; The behavior quantification subsystem includes a collection module, a matching module, a calculation module and a storage module; The collection module is used to collect the user's bank network consumption behavior data and obtain keywords in the transaction summary; The matching module is used to match the consumption theme corresponding to the consumption behavior according to the keywords; The calculation module is used to obtain the consumption behavior indicators under the consumption theme, and calculate the quantitative values ​​of each consumption behavior indicator according to the consumption behavior data; The storage module is used to store various consumption behavior indicators, quantitative values ​​and transaction time of the user; The abnormality identification and monitoring subsystem includes a comparison module, a judgment module, a marking module and a monitoring module; The comparison module is used to extract the historical quantitative value of the same consumption behavior indicator of the user and compare it with the new quantitative value of the consumption behavior indicator to obtain the change range and time interval of the quantitative value; The judgment module is used to judge whether the consumption behavior indicator of the user is abnormal according to the change range and time interval of the quantitative value; The marking module is used to mark users with abnormal consumption behavior; The monitoring module is used to monitor the marked users.

4. The user consumption behavior monitoring and analysis system based on network summary calculation according to claim 3 is characterized by: The anomaly identification and monitoring subsystem also includes a pattern acquisition module and a rule recording module; The pattern acquisition module is used to acquire the core behavior pattern of the user; the rule recording module is used to record the abnormal judgment rules and trigger rules; The abnormality judgment rules are as follows: General category: if the user's quantitative value changes beyond the set range, it is marked as abnormal; Investment category: If the change range of the user's quantitative value reaches the set value, it will be marked as abnormal; Qualification category: If the change range of the quantitative value of the user's consumption behavior indicator exceeds or is lower than N% of the quantitative value of the consumption behavior indicator of the group of users with the same core behavior pattern, it will be marked as abnormal; The triggering rules are as follows: General category: record the consumption tag of the user, and pass the aggregated user list and consumption tags to the relevant system for reference; Investment category: obtain user information and store it for use in marketing business; Qualification category: store user-related data and transmit it to the corresponding decision engine for corresponding rule verification; The judgment module is also used to judge whether the consumption behavior indicator of the user is abnormal according to the abnormal judgment rules and the core behavior pattern of the user; The monitoring module is used to monitor the marked users according to the core behavior patterns and triggering rules of the users.

Images