Insurance industry-oriented customer full life cycle event data modeling method and system

Abstract

The invention relates to an insurance industry-oriented customer full-life-cycle event data modeling method and system. The method comprises the steps of collecting customer full-life-cycle event data through a customer contact channel and a business processing link of an insurance industry; preprocessing the collected data, and converting the data into structured data; key elements are extracted from the preprocessed structured detail data; performing hierarchical and classified storage on the structured data based on a hierarchical and classified event data model, and respectively establishing a trunk index model and an expansion detail model in combination with the key elements; generating a unique event ID (Identity) for each event from the key elements; and performing data query optimization and providing data service based on the trunk index model, the expansion detail model and the event ID. Compared with the prior art, the method has the advantages that massive customer event data can be efficiently integrated and analyzed, the data processing and query capability is improved, more efficient and flexible data services are provided, and decision support is provided for the insurance industry.

Description

Technical Field

[0001] The present invention relates to the technical field of insurance industry data processing, and in particular to a customer full life cycle event data modeling method and system for the insurance industry. Background Art

[0002] As big data technologies such as Hadoop and MPP mature, the threshold for use is gradually decreasing, and there are more and more application scenarios in the insurance field. In particular, in risk assessment, customer service, customer marketing, etc., it is necessary to collect various types of customer business or behavior data and build a more refined customer portrait than the traditional model, thereby improving the accuracy and effectiveness of related models and services.

[0003] Building accurate customer portraits based on event data throughout the customer life cycle has always been the focus of applications in all walks of life. With the implementation of big data technology, this goal has become possible. Retaining and processing more complete and longer-term customer data means that more personalized services can be provided to different customers, improving marketing success rates, and comprehensively preventing potential risks. This is of great value to companies in reducing costs and increasing efficiency, improving service quality, and even improving brand competitiveness.

[0004] For national industry leaders, the number of customers reaches hundreds of millions. Even for the relatively low-frequency transaction model of the insurance industry, the amount of customer event data will be very large. Especially in the insurance field, the customer service cycle spans more than several decades. Faced with such a long period of massive data, how to effectively organize, extract and save the information contained in the data has become a very challenging issue. In this processing and statistical query scenario based on customer event data of more than 10 billion scales, traditional database technology has been difficult to meet the needs. Even if distributed database or big data technology is used, it is usually modeled for certain scenarios separately, and then the query indicators are generated by batch pre-calculation in a limited time period to reduce the amount of data processed and queried each time to meet the service performance requirements. However, this type of method still faces the dilemma of "blind men touching an elephant". Each time, only the customer event trajectory from a single perspective can be seen, and it is difficult to provide a full picture of events in the entire life cycle. At the same time, it cannot support flexible queries of customer trajectories and indicators based on the full amount of events. How to track the behavior trajectory of customers throughout their life cycle, trace back historical behavior, and build a more complete and accurate customer portrait is still an area worthy of further exploration. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a customer life cycle event data modeling method and system for the insurance industry, to improve the processing and query capabilities based on massive customer event detail data, and to provide more efficient and flexible data services.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A customer life cycle event data modeling method for the insurance industry includes the following steps:

[0008] Collect event data from the entire customer life cycle through the insurance industry’s customer contact channels and business processing links;

[0009] Preprocess the collected data and convert it into structured data;

[0010] Extract key elements from preprocessed structured detailed data;

[0011] Based on the hierarchical classification event data model, the structured data is hierarchically classified and stored, and a main index model and an extended detail model are established respectively in combination with the key elements;

[0012] Generating a unique event ID for each event in parallel from the key elements;

[0013] Data query optimization and data service provision are performed based on the main index model, extended detail model and event ID.

[0014] Furthermore, the event data collected throughout the customer's life cycle includes transaction records, application logs, and underwriting claim information;

[0015] The sources of the event data include tracking point logs and business transaction details. The tracking point logs use a common tracking point collection tool to record event data using a standardized tracking point log standard; the business transaction details come from event detail data recorded in various business systems;

[0016] The event data collection process includes:

[0017] Through data synchronization tools, real-time triggering or scheduled batch synchronization is used to aggregate tracking logs and business transaction details into the distributed file storage / message queue of the big data platform to complete the collection of event data.

[0018] Furthermore, the preprocessing operation includes cleaning, deduplication and formatting the collected data.

[0019] Furthermore, the preprocessing operation also includes parsing and cleaning the data of the tracking log using a unified tracking standard, extracting and converting each part into structured data, and formatting and converting the code values ​​of each field in the structured data according to unified technical specifications, so that the data content expression of the same meaning field is consistent.

[0020] Furthermore, the key elements include customer identification number, event time, event type, event name, event object and supplementary business information, and the supplementary business information includes sales channels, associated salespersons, affiliated institutions and associated insurance product information.

[0021] Furthermore, the hierarchical classification event data model includes a first-level event classification, each of which includes multiple second-level event classifications, and the first-level event classifications are divided into insurance policy classification, service rights classification, Internet online behavior classification, and traditional and offline behavior classification, and each first-level event classification contains common information of various types of events;

[0022] The secondary event classification includes specific information of sub-category events;

[0023] The secondary event classification and event-specific detailed information are saved in the corresponding data model through a standardized template for analysis when used;

[0024] For the event data of all customers, the key elements are integrated to form a backbone index model covering all key elements, and an extended detail model associated through the key elements is formed.

[0025] Furthermore, the generation of the event ID follows the principles of uniqueness and distributed parallelism.

[0026] Furthermore, the processing process of the data query optimization is specifically as follows:

[0027] The OLAP acceleration engine is used to optimize the model design for data query, and the optimization design includes:

[0028] Setting the customer number of the backbone index model as the distribution key for the query so that the backbone index model is associated with the customer dimension table and retrieved using local association;

[0029] Set the event ID as the distribution key of the extended detail model to locate the event data;

[0030] Set the event time as the segment index so that the data is distributed in the physical file in an orderly manner;

[0031] Set the customer number as the clustered index so that the data blocks in the file are clustered by customer number;

[0032] Set the event type as a low-cardinality column bitmap index to query for specific event types.

[0033] Furthermore, the data service includes online query and batch synchronization. The online query includes point query and analytical query. The batch synchronization is used to provide data for algorithm model training to achieve customer behavior portrait prediction and business opportunity lead discovery based on the trained algorithm model and according to future single or time window behaviors.

[0034] The present invention also provides a customer life cycle event data modeling system for the insurance industry, comprising a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method described above.

[0035] Compared with the prior art, the present invention has the following advantages:

[0036] (1) The present invention first realizes the collection of customer life cycle event data by comprehensively using various technical components of big data. In view of the characteristics of customer life cycle event data in the insurance industry, the present invention abstracts key business elements and designs a data modeling method of a three-level classification event system detail model + a backbone index model, thereby realizing the integration of full event data and performing targeted index optimization for common application scenarios.

[0037] On the whole, the present invention can conveniently and quickly retrieve all customer events and calculate statistical indicators through the backbone index model, and can also quickly locate the classification event model to obtain event details; it gives full play to the computing power of large-scale distributed clusters, effectively improves the processing and query capabilities based on massive customer event detail data, and provides more efficient and flexible data services.

[0038] (2) Comprehensively utilize big data technology to achieve horizontal scalability and improve service performance:

[0039] The present invention comprehensively utilizes the large-scale distributed parallel processing mechanism technology components of big data technology, including distributed storage, computing engine, OLAP engine, etc., to maximize the use of software and hardware resources, have horizontal scalability, and improve service performance.

[0040] (3) The present invention provides data key element identification, a three-level classification event data model, and a corresponding query optimization method based on the characteristics of the insurance industry, which can adapt to the characteristics of the insurance industry with a large number of existing customers and a long time span.

[0041] (4) The processing method of each link of the present invention is not bound to specific technical components or targeted at specific scenarios. It not only supports batch data processing, but is also applicable to real-time streaming data processing. In similar massive behavioral data processing and service scenarios, the modeling method and data processing process can be reused. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1A flow chart of a customer life cycle event data modeling method for the insurance industry provided in an embodiment of the present invention;

[0043] Figure 2 This is a diagram showing an application example of a customer life cycle event data modeling method for the insurance industry provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0045] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0047] Example 1

[0048] like Figure 1 As shown, this embodiment provides a customer life cycle event data modeling method for the insurance industry, including the following steps:

[0049] Data collection step S1: Collect event data of customers throughout their life cycle through the insurance industry’s customer contact channels and business processing links;

[0050] Data preprocessing step S2: preprocessing the collected data and converting it into structured data;

[0051] Key element extraction step S3: extracting key elements from the preprocessed structured detailed data;

[0052] Data modeling and integration step S4: hierarchical classification and storage of structured data based on the hierarchical classification event data model, and establishing a backbone index model and an extended detail model in combination with key elements;

[0053] Event ID generation step S5: Generate a unique event ID for each event in parallel from the key elements;

[0054] Query optimization step S6: using the OLAP acceleration engine to optimize data query based on the main index model, the extended detail model and the event ID;

[0055] Data service step S7: Based on the established data model, data services are provided for application scenarios.

[0056] The following is a detailed description of each step:

[0057] 1. Data collection steps

[0058] Specifically, the event data collected throughout the customer life cycle includes, but is not limited to, transaction records, application logs, and underwriting claims information;

[0059] From the technical implementation point of view, there are two main methods: tracking log and business transaction details. Among them, tracking log uses a general tracking collection tool to standardize the definition of tracking log standards, record user information, event type, time, object and other supplementary business information. Business transaction details come from the event detail data recorded in each business system and are generated as the business occurs.

[0060] The process of collecting event data includes:

[0061] Through data synchronization tools, real-time triggering or scheduled batch synchronization is used to aggregate tracking logs and business transaction details into the distributed file storage / message queue of the big data platform to complete the collection of original customer event data.

[0062] 2. Data preprocessing steps

[0063] The collected data is pre-processed by cleaning, deduplication, formatting and other operations to ensure data quality.

[0064] For embedded point log data, the data parsing and cleaning process can be solidified by defining a unified embedded point standard, and each part can be extracted and converted into structured data. Then, the relevant fields are formatted and converted into code values ​​according to unified technical specifications, so that the data content of the fields with the same meaning is expressed consistently.

[0065] 3. Key element extraction steps

[0066] Extract key elements from preprocessed structured detail data.

[0067] In the customer life cycle event data, identify and define key elements: customer identification number, event time, event type, event name, event object and supplementary business information. Among them, the supplementary business information focuses on identifying the sales channels, related salesmen, affiliated institutions, and related insurance product information that the insurance business is concerned about.

[0068] 4. Data modeling and integration steps

[0069] The hierarchical classification event data model includes a first-level event classification, each of which includes multiple second-level event classifications. The first-level event classification is divided into insurance policy classification, service rights classification, Internet online behavior classification, and traditional and offline behavior classification. Each first-level event classification contains common information for various types of events.

[0070] The second-level event classification contains the specific information of the sub-category event;

[0071] The secondary event classification and event-specific detailed information are saved in the corresponding data model through a standardized template for analysis when used;

[0072] For all customers' event data, key elements are integrated to form a backbone index model covering all key elements, and an extended detail model is formed that is associated through key elements.

[0073] Specifically, this step refers to the dimensional modeling theory and adopts the divide-and-conquer approach to establish a unified hierarchical classification event data model that supports multiple query usage scenarios. In the context of the insurance industry, the customer's full life cycle event data is abstracted into a three-level classification event system detailed model: first-level event categories, second-level event subcategories, and third-level events. Third-level events can be increased as needed.

[0074] The first-level event categories are divided into four categories: insurance policy, service rights, Internet online behavior, and traditional and offline behavior. Each first-level classification contains common information of various events. For example, insurance policy events are usually associated with insurance policies and salesperson information, and Internet online behavior includes Internet application information of customers. The first-level classification includes several second-level event subcategories, and each subcategory contains unique information of the subcategory event. For example, the insurance second-level event subcategory under the first-level event category of insurance policy includes insurance policy information, underwriting status, insured amount, insured person information, etc.; complaint events in the traditional and offline behavior category include complaint number, complainee information, channel, type, content and processing progress, etc. Specific explanations are as follows:

[0075] 1) Insurance policy: customer business events related to the underwriting, preservation, and claims settlement of insurance policies, such as insurance application, renewal, preservation application, case reporting, payment, policy cancellation, and policy loan.

[0076] 2) Service and rights category: events related to services and rights provided to customers, such as customer consultation, complaints, rights collection, activation, and use.

[0077] 3) Online Internet Behavior: Customer behaviors collected from online customer touchpoints, such as APP, WeChat public account or mini program, corporate WeChat, etc. Such as customer registration, browsing, following, corporate WeChat chat, article subscription, sharing, mini program activity participation, etc.

[0078] 4) Traditional and offline behavior: customer behaviors recorded and collected through traditional customer service channels or offline marketing activities, such as appointment queues at outlets, SMS calls, offline visits by salesmen, and participation in theme activities.

[0079] Establish independent customer event data models for the above four types of first-level events, store event data belonging to this major category, and save their common information. The subcategories of second-level events and the unique detailed information of the events are saved in independent fields of their respective models by defining standardized json templates for parsing when used. Then establish a full detailed table containing all customer events. That is, for all customer event detail data, the key elements extracted in the above step S3 are integrated to form a backbone index model covering all key elements of the events, and the key element information therein is quickly associated with the four types of extended details or public dimension tables.

[0080] The integration of the above data models can be done in real-time or batch incremental integration, which is fastest and consistent with the frequency of data collection. It can also be done in a unified fixed frequency integration to simplify the processing flow.

[0081] 5. Event ID generation steps

[0082] Generate a unique event ID for each event so that the required detailed data can be retrieved in the backbone or extended data model. The generation of event ID follows the principles of uniqueness and distributed parallelism. For example, using MD5 (customer number, event time, event category, event name, event object) can meet the above two requirements and achieve fast unique ID generation.

[0083] 6. Query optimization steps

[0084] The specific process of data query optimization is as follows:

[0085] Use OLAP acceleration engines such as MPP architecture to optimize model design for data query.

[0086] That is, using the OLAP acceleration engine, targeting specific data service scenarios, and optimizing the design of physical models. This is a key step to improve the query service performance of massive data based on big data technology. For example, in the common scenario of tracing back the event details and related extended attributes of a specific time period in the entire life cycle of a single customer, but not returning all data columns, column storage technology can be used to avoid scanning irrelevant data.

[0087] Optimized design includes:

[0088] Set the customer number of the backbone index model as the distribution key for the query, so that the backbone index model data is scattered and distributed by customer number in the distributed system. This ensures that local association is used when the backbone index model is associated with the customer dimension table for retrieval, thus avoiding the network transmission of large amounts of data across nodes.

[0089] Set the event ID as the distribution key of the extended detail model (if all information of a single event needs to be queried each time, use row storage) to quickly locate event data;

[0090] Set event time as the segment index to ensure the orderly distribution of data in physical files. Use bloom filtering and other technologies to quickly exclude as many impossible files as possible during query, thus improving the efficiency of event-time-based queries.

[0091] Set the customer number as a clustered index so that the data blocks in the file are clustered by customer number, improving the performance of equal value queries by number;

[0092] Set the event type as a low-cardinality column bitmap index to further improve the query efficiency for specific event types.

[0093] 7. Data service steps

[0094] Data services include online query and batch synchronization. Online query includes point query and analytical query, such as event trajectory query of a single customer or aggregate statistical query based on time window as analytical query, and query of details of a single event of a single customer as point query. Batch synchronization is used to provide data for algorithm model training, so as to realize customer behavior profile prediction and business opportunity clue discovery based on the trained algorithm model and future single transaction or time window behavior.

[0095] Based on the event model generated in step S4 and the physical model optimization in step S6, corresponding scenario services are provided to meet performance requirements.

[0096] The following is a specific solution application example:

[0097] like Figure 2 As shown in the figure, this example uses Alibaba's big data MaxCompute batch computing engine, DataWorks integrated development platform and scheduling system, and Hologres database (OLAP engine) to clean and model data and provide query services. The specific implementation steps are as follows:

[0098] 1. Data collection: Use the DataWorks batch data synchronization tool to extract business system data, and use point-of-care log collection tools such as Sensors to collect behavioral log data from the customer service system.

[0099] 2. Data preprocessing: Based on the big data platform and technical specifications, use MaxCompute to complete data integration and normalization processing.

[0100] 3. Extraction of key elements: Based on the characteristics of the insurance industry, identify and extract key elements as the main content of the backbone model.

[0101] 4. Data modeling and integration: By classifying and abstracting the four major types of customer events in the insurance industry, we establish a backbone index model and an extended detail model.

[0102] 5. Event ID generation: From the key elements, a unique event ID is generated in parallel as the association key.

[0103] 6. Query optimization: Synchronize data to the OLAP engine Hologres database, optimize physical characteristics for specific data service scenarios, and improve query performance.

[0104] 7. Data service: Build data synchronization or interface query services for algorithm models or query services.

[0105] This embodiment also provides a customer life cycle event data modeling system for the insurance industry, including a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the customer life cycle event data modeling method for the insurance industry as described above.

[0106] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A customer life cycle event data modeling method for the insurance industry, characterized in that: The following steps are involved: Collect event data from the entire customer life cycle through the insurance industry’s customer contact channels and business processing links; Preprocess the collected data and convert it into structured data; Extract key elements from preprocessed structured detailed data; Based on the hierarchical classification event data model, the structured data is hierarchically classified and stored, and a main index model and an extended detail model are established respectively in combination with the key elements; Generating a unique event ID for each event in parallel from the key elements; Data query optimization and data service provision are performed based on the main index model, extended detail model and event ID.

2. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The event data collected throughout the customer's life cycle includes transaction records, application logs, and underwriting claims information; The sources of the event data include tracking logs and business transaction details. The tracking logs use a common tracking collection tool to record event data using a standardized tracking log standard. The business transaction details come from event detail data recorded in various business systems; The event data collection process includes: Through data synchronization tools, real-time triggering or scheduled batch synchronization is used to aggregate tracking logs and business transaction details into the distributed file storage / message queue of the big data platform to complete the collection of event data.

3. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The preprocessing operation includes cleaning, deduplication and formatting the collected data.

4. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The preprocessing operation also includes parsing and cleaning the data of the tracking log using a unified tracking standard, extracting and converting each part into structured data, and formatting and converting the code values ​​of each field in the structured data according to unified technical specifications, so that the data content of the field with the same meaning is expressed consistently.

5. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The key elements include customer identification number, event time, event type, event name, event object and supplementary business information. The supplementary business information includes sales channels, associated salespersons, affiliated institutions and associated insurance product information.

6. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The hierarchical classification event data model includes a first-level event classification, each of which includes multiple second-level event classifications. The first-level event classifications are divided into insurance policy classifications, service rights classifications, Internet online behavior classifications, and traditional and offline behavior classifications. Each first-level event classification contains common information for each type of event. The secondary event classification includes specific information of sub-category events; The secondary event classification and event-specific detailed information are saved in the corresponding data model through a standardized template for analysis when used; For the event data of all customers, the key elements are integrated to form a backbone index model covering all key elements, and an extended detail model associated through the key elements is formed.

7. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The generation of the event ID follows the principles of uniqueness and distributed parallelism.

8. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The processing process of data query optimization is specifically as follows: The OLAP acceleration engine is used to optimize the model design for data query, and the optimization design includes: Setting the customer number of the backbone index model as the distribution key for the query so that the backbone index model is associated with the customer dimension table and retrieved using local association; Set the event ID as the distribution key of the extended detail model to locate the event data; Set the event time as the segment index so that the data is distributed in the physical file in an orderly manner; Set the customer number as the clustered index so that the data blocks in the file are clustered by customer number; Set the event type as a low-cardinality column bitmap index to query for specific event types.

9. A customer life cycle event data modeling method for the insurance industry according to claim 1, characterized in that: The data service includes online query and batch synchronization. The online query includes point query and analytical query. The batch synchronization is used to provide data for algorithm model training to achieve customer behavior portrait prediction and business opportunity lead discovery based on the trained algorithm model and future single or time window behavior.

10. A customer life cycle event data modeling system for the insurance industry, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of any one of the methods according to claims 1 to 9.

Images