Precise financial customer portrait analysis platform based on big data

Abstract

The invention discloses a precise financial customer portrait analysis platform based on big data, which aims at the defects of the traditional financial customer analysis means, and comprises a plurality of key modules: a self-adaptive data acquisition module which is provided with an intelligent sensing interface, can automatically adjust an acquisition strategy according to the change of a data source, fuses multi-modal data, and acquires the image of the financial customer through the self-adaptive data acquisition module; the block chain is used for ensuring the reliability and safety of the data; according to the hyper-fusion data storage module, heterogeneous data integrated storage is combined with intelligent layering, and efficient query and low-cost storage are achieved; an intelligent driving data analysis module, an adaptive algorithm recommendation engine on-demand matching algorithm and small sample learning solve sample insufficiency, and cooperate with a dual-mode pipeline to guarantee analysis timeliness and depth; the dynamic evolution portrait construction module is used for updating portraits in real time, customizing personalized tags and comparing reference groups; and a security collaborative application interface module and a multi-tenant isolation interface are matched to realize bidirectional ecological feedback. The platform can accurately draw customer portraits, enable financial businesses and improve the competitiveness of organizations.

Description

Technical Field

[0001] The present invention belongs to the technical field related to financial data analysis, and specifically relates to a precise financial customer portrait analysis platform based on big data. Background Art

[0002] With the booming financial market and the accelerated digitalization process, financial institutions are facing an increasingly complex and diverse customer base, with increasingly personalized needs. Traditional financial customer analysis methods, limited by a single data source and insufficient processing capabilities, are often limited to crude customer classification and superficial understanding based on limited internal bank structured data, such as basic account information and simple transaction records. This approach fails to capture customers' potential financial needs, credit characteristics, and consumption preferences as they manifest across a full range of scenarios, including daily life, online consumption, and social interactions. In today's era of highly integrated internet finance, massive amounts of unstructured data, including social media activity, e-commerce purchases, and online browsing history, contain a wealth of customer insights, yet they are not effectively integrated and utilized. Existing customer analysis tools either focus on mining a single type of data or encounter challenges with heterogeneous data when integrating data. This leads to significant bias in analytical results and low profiling accuracy, making them unable to provide solid support for key business processes such as precision marketing, risk management, and personalized service customization, significantly hindering their market expansion and competitiveness. Summary of the Invention

[0003] The purpose of the present invention is to provide a precise financial customer profile analysis platform based on big data to solve the problems raised in the above background technology.

[0004] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions:.

[0005] A precise financial customer profile analysis platform based on big data, including:

[0006] Adaptive data acquisition module:

[0007] Configure an intelligent sensing data collection interface to dynamically identify data source types, data update frequency, and data format changes, and automatically adapt and adjust collection strategies to ensure continuous and complete data collection from financial institutions' internal systems and external data sources;

[0008] It has a multimodal data fusion acquisition component that can simultaneously capture unstructured data in different modalities such as text, images, audio, and video, and extract financial-related information based on predefined financial semantic rules;

[0009] Use blockchain technology to trace and verify some key data sources, and use blockchain encryption features to ensure data security and prevent data tampering during data collection and transmission;

[0010] Hyper-converged data storage module:

[0011] Build an integrated storage architecture for heterogeneous data. By designing a data index conversion layer, based on the separation of underlying physical storage, enable upper-layer applications to seamlessly query and analyze structured, semi-structured, and unstructured data.

[0012] An intelligent hot and cold data tiering management mechanism is introduced. Based on data access popularity, timeliness, and business relevance, data is automatically divided into hot, warm, and cold data areas, each using different storage media and optimization strategies.

[0013] Intelligent drive data analysis module:

[0014] It has an adaptive algorithm recommendation engine that combines financial business scenarios with data characteristics to automatically screen and match the optimal machine learning and deep learning algorithm combinations without manual intervention, enabling accurate analysis of data at different stages and complexities.

[0015] A small sample learning enhancement submodule uses transfer learning technology to introduce external large sample data from similar business scenarios for pre-training, and then fine-tunes with a small number of internal target samples to address the problem of insufficient sample data from rare customers and improve the model's generalization ability and accurate prediction results on small samples.

[0016] The integrated real-time stream computing and batch processing dual-mode analysis pipeline responds instantly to hot data such as real-time financial transactions and social dynamics based on event triggering mechanisms, updating key customer profile indicators in real time. It also uses batch processing to conduct periodic in-depth reviews of historical big data to uncover long-term customer behavior patterns.

[0017] Dynamic evolution portrait construction module:

[0018] Establish a real-time dynamic customer profile update mechanism. Relying on real-time data analysis results, once key customer behavior changes are detected, the profile update process is immediately triggered to ensure that the profile always stays close to the customer's actual status;

[0019] Develop an automatic generation algorithm for a personalized labeling system to extract customized labels from massive analysis results based on the needs of different financial business lines, thereby improving the targeted nature of the profiling.

[0020] Introducing cross-customer group comparative analysis capabilities to compare target customers with multi-dimensional reference customer groups of the same type, region, and income range, presenting customers' unique strengths and potential weaknesses in relative terms.

[0021] Security collaboration application interface module:

[0022] Design a multi-tenant secure isolation application interface. Based on tenant permission configuration, implement strict isolation of customer profile data throughout the entire storage, transmission, and access process, preventing data leakage across boundaries and ensuring the data independence and privacy security of each business unit.

[0023] Building a two-way feedback API ecosystem not only allows the internal business systems of financial institutions to call portrait data, but also allows the feedback data of the business systems during the actual use of portraits to be transmitted back to the platform. The platform continuously optimizes the portrait analysis algorithm and model parameters based on the feedback.

[0024] Preferably, in the adaptive data collection module, when the intelligent perception data collection interface detects that a new field is added or the data type is changed in the internal system database table structure of a financial institution, the change is detected through the pre-implanted metadata comparison module, and the ETL process configuration file is automatically adjusted to ensure that the new data is collected accurately; for external emerging social platforms, the interface uses the dynamic configuration function of the web crawler to update the crawling rules in real time according to changes in the platform page structure.

[0025] Preferably, in the adaptive data acquisition module, when the multimodal data fusion acquisition component processes the social platform information of pictures shared by customers, the image recognition module combines with the financial knowledge base to identify high-end consumption scenes in the pictures, the speech-to-text module converts customer-related audio into text, and the text analysis sub-module extracts financial keywords, integrating different modal information to provide material for subsequent customer portrait construction.

[0026] Preferably, in the hyper-converged data storage module, the intermediate data index conversion layer of the heterogeneous data integrated storage architecture first parses the query statement when processing the upper-level application query, splits it into sub-queries for different types of data, and then summarizes the results of each sub-query and returns them to the upper-level application in a unified format.

[0027] Preferably, in the hyper-converged data storage module, the intelligent hot and cold data tiered management mechanism runs the data heat evaluation algorithm on a daily basis, comprehensively considering factors such as data access frequency, last access time, and business-related weight, and scores and sorts various types of data. Hot data with scores higher than the threshold are stored in a high-performance SSD storage array, and a cache optimization strategy is used to ensure high-frequency reading and writing.

[0028] Preferably, in the intelligent-driven data analysis module, the adaptive algorithm recommendation engine selects a decision tree classification model based on the small amount of information and business scenarios initially entered by the customer at the beginning of a new customer account opening. As the customer's subsequent data accumulates, the model switching process is automatically triggered, and the deep neural network model is introduced for retraining and optimization.

[0029] Preferably, in the intelligent-driven data analysis module, when processing the credit risk assessment of ultra-high net worth customers, the small sample learning enhancement sub-module obtains a large amount of similar high net worth customer credit data from external cooperative financial institutions for pre-training, and then combines it with a small amount of ultra-high net worth customer sample data within the bank to fine-tune the model's specific layer parameters based on transfer learning.

[0030] Preferably, in the dynamic evolution portrait construction module, the real-time dynamic customer portrait update mechanism immediately initiates the portrait update process when it monitors the key behavioral changes of the customer repaying the mortgage in advance and transferring the funds into the stock account, re-evaluates the characteristics of the customer's capital liquidity and investment risk preference through the data analysis module, updates the portrait-related tags, and pushes them to the corresponding business department.

[0031] Preferably, in the dynamic evolution portrait construction module, the personalized label system automatically generates an algorithm in the wealth management business, which generates labels such as "conservative saver" and "aggressive investor" through data cluster analysis of customer investment product holding records, transaction frequency, and return preferences.

[0032] Preferably, in the dynamic evolution portrait construction module, the cross-customer group comparative analysis function selects local customer groups of the same industry and similar-sized enterprises as a reference when the bank expands its small and medium-sized enterprise loan business, and uses a radar chart to intuitively present the relative advantages and disadvantages of the target customers.

[0033] Compared with the existing technology, this invention provides a precise financial customer profile analysis platform based on big data, which has the following beneficial effects:

[0034] This invention, with its adaptive collection technology, greatly broadens the breadth and depth of data acquisition, accurately captures ever-changing customer information, and solves the problems of delayed and one-sided traditional data collection, allowing financial institutions to always grasp the latest customer dynamics and lay a solid foundation for subsequent accurate profiling.

[0035] The hyper-converged storage architecture breaks down data silos, offering integrated retrieval and intelligent tiered management. This significantly reduces storage costs while improving data utilization efficiency, providing strong support for efficient analysis of large data volumes.

[0036] Intelligently driven data analysis flexibly adapts to business needs and data characteristics. Small sample learning enhancement and a dual-mode computing pipeline enable the model to maintain high accuracy and strong adaptability in complex and changing financial scenarios, enabling precise responses to new customer acquisition and high-risk warnings.

[0037] Dynamically evolving profiles reflect customer changes, personalized tags, and cross-group comparisons in real time, helping financial institutions conduct targeted marketing and risk management, deeply tap into customer value, and significantly improve customer service experience and business conversion rates.

[0038] Under the premise of ensuring data security, the secure collaborative interface builds a two-way feedback loop to promote the collaborative progress of the platform and business systems, continuously improve the quality of portraits, and strengthen the overall operational collaboration efficiency of financial institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is the overall architecture diagram of the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] The present invention provides Figure 1 The illustrated platform is a precise financial customer profile analysis platform based on big data, including:

[0042] Adaptive data acquisition module:

[0043] Configure an intelligent sensing data collection interface to dynamically identify data source types, data update frequency, and data format changes, and automatically adapt and adjust collection strategies to ensure continuous and complete data collection from financial institutions' internal systems and external data sources;

[0044] It has a multimodal data fusion acquisition component that can simultaneously capture unstructured data in different modalities such as text, images, audio, and video, and extract financial-related information based on predefined financial semantic rules;

[0045] Use blockchain technology to trace and verify some key data sources, and use blockchain encryption features to ensure data security and prevent data tampering during data collection and transmission;

[0046] Hyper-converged data storage module:

[0047] Build an integrated storage architecture for heterogeneous data. By designing a data index conversion layer, based on the separation of underlying physical storage, enable upper-layer applications to seamlessly query and analyze structured, semi-structured, and unstructured data.

[0048] An intelligent hot and cold data tiering management mechanism is introduced. Based on data access popularity, timeliness, and business relevance, data is automatically divided into hot, warm, and cold data areas, each using different storage media and optimization strategies.

[0049] Intelligent drive data analysis module:

[0050] It has an adaptive algorithm recommendation engine that combines financial business scenarios with data characteristics to automatically screen and match the optimal machine learning and deep learning algorithm combinations without manual intervention, enabling accurate analysis of data at different stages and complexities.

[0051] A small sample learning enhancement submodule uses transfer learning technology to introduce external large sample data from similar business scenarios for pre-training, and then fine-tunes with a small number of internal target samples to address the problem of insufficient sample data from rare customers and improve the model's generalization ability and accurate prediction results on small samples.

[0052] The integrated real-time stream computing and batch processing dual-mode analysis pipeline responds instantly to hot data such as real-time financial transactions and social dynamics based on event triggering mechanisms, updating key customer profile indicators in real time. It also uses batch processing to conduct periodic in-depth reviews of historical big data to uncover long-term customer behavior patterns.

[0053] Dynamic evolution portrait construction module:

[0054] Establish a real-time dynamic customer profile update mechanism. Relying on real-time data analysis results, once key customer behavior changes are detected, the profile update process is immediately triggered to ensure that the profile always stays close to the customer's actual status;

[0055] Develop an automatic generation algorithm for a personalized labeling system to extract customized labels from massive analysis results based on the needs of different financial business lines, thereby improving the targeted nature of the profiling.

[0056] Introducing cross-customer group comparative analysis capabilities to compare target customers with multi-dimensional reference customer groups of the same type, region, and income range, presenting customers' unique strengths and potential weaknesses in relative terms.

[0057] Security collaboration application interface module:

[0058] Design a multi-tenant secure isolation application interface. Based on tenant permission configuration, implement strict isolation of customer profile data throughout the entire storage, transmission, and access process, preventing data leakage across boundaries and ensuring the data independence and privacy security of each business unit.

[0059] Building a two-way feedback API ecosystem not only allows the internal business systems of financial institutions to call portrait data, but also allows the feedback data of the business systems during the actual use of portraits to be transmitted back to the platform. The platform continuously optimizes the portrait analysis algorithm and model parameters based on the feedback.

[0060] In the adaptive data collection module, when the intelligent perception data collection interface detects the addition of new fields or changes in data types in the internal system database table structure of a financial institution, it detects the changes through the pre-implanted metadata comparison module and automatically adjusts the ETL process configuration file to ensure that the new data is collected accurately. For external emerging social platforms, the interface uses the dynamic configuration function of the web crawler to update the crawling rules in real time according to changes in the platform page structure.

[0061] In the adaptive data collection module, when the multimodal data fusion collection component processes the social platform information of pictures shared by customers, the image recognition module combines with the financial knowledge base to identify high-end consumption scenarios in the pictures, the speech-to-text module converts customer-related audio into text, and the text analysis sub-module extracts financial keywords, integrating different modal information to provide material for subsequent customer portrait construction.

[0062] In the hyper-converged data storage module, the intermediate data index conversion layer of the heterogeneous data integrated storage architecture first parses the query statement when processing the upper-level application query, splits it into sub-queries for different types of data, and then summarizes the results of each sub-query and returns them to the upper-level application in a unified format.

[0063] In the hyper-converged data storage module, the intelligent hot and cold data tiered management mechanism runs a data heat assessment algorithm on a daily basis, comprehensively considering factors such as data access frequency, last access time, and business-related weight, and scores and sorts various types of data. Hot data with scores higher than the threshold is stored in a high-performance SSD storage array, and a cache optimization strategy is used to ensure high-frequency reading and writing.

[0064] In the intelligent-driven data analysis module, the adaptive algorithm recommendation engine selects a decision tree classification model based on the small amount of information and business scenarios initially entered by the customer at the beginning of a new customer account opening. As the customer's subsequent data accumulates, the model switching process is automatically triggered, and the deep neural network model is introduced for retraining and optimization.

[0065] In the intelligent-driven data analysis module, when processing the credit risk assessment of ultra-high net worth customers, the small sample learning enhancement sub-module obtains a large amount of similar high-net-worth customer credit data from external cooperative financial institutions for pre-training, and then combines it with a small amount of ultra-high net worth customer sample data within the bank to fine-tune the parameters of the model's specific layers based on transfer learning.

[0066] In the dynamic evolution portrait construction module, the real-time dynamic customer portrait update mechanism immediately initiates the portrait update process when it detects key behavioral changes such as customers repaying their mortgages in advance and transferring funds to their stock accounts. It re-evaluates the customer's liquidity and investment risk preference characteristics through the data analysis module, updates the relevant portrait tags, and pushes them to the corresponding business departments.

[0067] In the dynamic evolution portrait construction module, the personalized label system automatically generates an algorithm in the wealth management business. Through data cluster analysis of customers' investment product holding records, transaction frequency, and return preferences, it generates labels such as "conservative savers" and "aggressive investors."

[0068] In the dynamic evolution portrait construction module, the cross-customer group comparative analysis function selects local customer groups of the same industry and similar size as a reference when banks expand their small and medium-sized enterprise loan business, and uses radar charts to intuitively present the relative advantages and disadvantages of target customers.

[0069] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein. Any modification, replacement, and improvement made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A precise financial customer profile analysis platform based on big data, characterized by: include: Adaptive data acquisition module: Configure an intelligent sensing data collection interface to dynamically identify data source types, data update frequency, and data format changes, and automatically adapt and adjust collection strategies to ensure continuous and complete data collection from financial institutions' internal systems and external data sources; It has a multimodal data fusion acquisition component that can simultaneously capture unstructured data in different modalities such as text, images, audio, and video, and extract financial-related information based on predefined financial semantic rules; Use blockchain technology to trace and verify some key data sources, and use blockchain encryption features to ensure data security and prevent data tampering during data collection and transmission; Hyper-converged data storage module: Build an integrated storage architecture for heterogeneous data. By designing a data index conversion layer, based on the separation of underlying physical storage, enable upper-layer applications to seamlessly query and analyze structured, semi-structured, and unstructured data. An intelligent hot and cold data tiering management mechanism is introduced. Based on data access popularity, timeliness, and business relevance, data is automatically divided into hot, warm, and cold data areas, each using different storage media and optimization strategies. Intelligent drive data analysis module: It has an adaptive algorithm recommendation engine that combines financial business scenarios with data characteristics to automatically screen and match the optimal machine learning and deep learning algorithm combinations without manual intervention, enabling accurate analysis of data at different stages and complexities. A small sample learning enhancement submodule uses transfer learning technology to introduce external large sample data from similar business scenarios for pre-training, and then fine-tunes with a small number of internal target samples to address the problem of insufficient sample data from rare customers and improve the model's generalization ability and accurate prediction results on small samples. The integrated real-time stream computing and batch processing dual-mode analysis pipeline responds instantly to hot data such as real-time financial transactions and social dynamics based on event triggering mechanisms, updating key customer profile indicators in real time. It also uses batch processing to conduct periodic in-depth reviews of historical big data to uncover long-term customer behavior patterns. Dynamic evolution portrait construction module: Establish a real-time dynamic customer profile update mechanism. Relying on real-time data analysis results, once key customer behavior changes are detected, the profile update process is immediately triggered to ensure that the profile always stays close to the customer's actual status; Develop an automatic generation algorithm for a personalized labeling system to extract customized labels from massive analysis results based on the needs of different financial business lines, thereby improving the targeted nature of the profiling. Introducing cross-customer group comparative analysis capabilities to compare target customers with multi-dimensional reference customer groups of the same type, region, and income range, presenting customers' unique strengths and potential weaknesses in relative terms. Security collaboration application interface module: Design a multi-tenant secure isolation application interface. Based on tenant permission configuration, implement strict isolation of customer profile data throughout the entire storage, transmission, and access process, preventing data leakage across boundaries and ensuring the data independence and privacy security of each business unit. Building a two-way feedback API ecosystem not only allows the internal business systems of financial institutions to call portrait data, but also allows the feedback data of the business systems during the actual use of portraits to be transmitted back to the platform. The platform continuously optimizes the portrait analysis algorithm and model parameters based on the feedback.

2. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the adaptive data collection module, when the intelligent perception data collection interface detects that a new field is added or the data type is changed in the internal system database table structure of a financial institution, it detects the change through the pre-implanted metadata comparison module and automatically adjusts the ETL process configuration file to ensure that the new data is collected accurately; for external emerging social platforms, the interface uses the dynamic configuration function of the web crawler to update the crawling rules in real time according to changes in the platform page structure.

3. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the adaptive data acquisition module, when the multimodal data fusion acquisition component processes the social platform information of pictures shared by customers, the image recognition module combines with the financial knowledge base to identify high-end consumption scenarios in the pictures, the speech-to-text module converts customer-related audio into text, and the text analysis sub-module extracts financial keywords, integrating different modal information to provide material for subsequent customer portrait construction.

4. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the hyper-converged data storage module, the intermediate data index conversion layer of the heterogeneous data integrated storage architecture first parses the query statement when processing the upper-level application query, splits it into sub-queries for different types of data, and then summarizes the results of each sub-query and returns them to the upper-level application in a unified format.

5. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the hyper-converged data storage module, the intelligent hot and cold data tiered management mechanism runs a data heat assessment algorithm on a daily basis, comprehensively considering factors such as data access frequency, last access time, and business-related weight, and scores and sorts various types of data. Hot data with scores higher than the threshold is stored in a high-performance SSD storage array, and a cache optimization strategy is used to ensure high-frequency reading and writing.

6. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the intelligent-driven data analysis module, the adaptive algorithm recommendation engine selects a decision tree classification model based on the small amount of information and business scenarios initially entered by the customer at the beginning of a new customer account opening. As the customer's subsequent data accumulates, the model switching process is automatically triggered, and the deep neural network model is introduced for retraining and optimization.

7. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the intelligent-driven data analysis module, when processing the credit risk assessment of ultra-high net worth customers, the small sample learning enhancement sub-module obtains a large amount of similar high-net-worth customer credit data from external cooperative financial institutions for pre-training, and then combines it with a small amount of ultra-high net worth customer sample data within the bank to fine-tune the parameters of the model's specific layers based on transfer learning.

8. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the dynamic evolution portrait construction module, the real-time dynamic customer portrait update mechanism immediately initiates the portrait update process when it detects key behavioral changes such as customers repaying their mortgages in advance and transferring funds to their stock accounts. It re-evaluates the characteristics of the customer's liquidity and investment risk preferences through the data analysis module, updates the relevant labels of the portrait, and pushes them to the corresponding business departments.

9. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the dynamic evolution portrait construction module, the personalized label system automatically generates an algorithm in the wealth management business, which generates labels such as "conservative saver" and "aggressive investor" through data cluster analysis of customers' investment product holding records, transaction frequency, and return preferences.

10. The big data-based precise financial customer profile analysis platform according to claim 1, characterized in that: In the dynamic evolution portrait construction module, the cross-customer group comparative analysis function selects local customer groups of the same industry and similar size as a reference when banks expand their small and medium-sized enterprise loan business, and uses radar charts to intuitively present the relative advantages and disadvantages of target customers.

Images