CN-120807087-B - Health risk automatic quotation system based on big data

CN120807087BCN 120807087 BCN120807087 BCN 120807087BCN-120807087-B

Abstract

The invention relates to the technical field of big data, in particular to a health risk automatic quotation system based on big data, wherein the invention integrates sleeping depth, health behaviors and living environment data in real time through a multi-source heterogeneous data acquisition module, so as to solve the problem of single dimension of traditional health risk data; the dynamic health portrait construction module adopts a dynamic weight distribution and space-time alignment technology, unifies multi-granularity data into weighted daily dimension vectors, eliminates data value evaluation differences and time granularity cracks, generates real-time dynamic health risk scores based on the weighted daily dimension vectors, extracts health behavior entropy quantification behavior stability, breaks through static risk evaluation limitation, and constructs a personalized risk evolution path diagram through entropy abnormal duration identification and disease development rate mapping by the risk conduction prediction module, finally converts risk upgrading nodes into accurate factors, dynamically generates stepped premium gradients, and realizes accurate matching of premium level and user real-time risk tracks.

Inventors

LIU JUN
LI XINGJIAN
MA JI
WU HONGFENG
ZHONG JUNFANG
AN SHENGZHI

Assignees

中国人民健康保险股份有限公司深圳分公司

Dates

Publication Date: 20260508
Application Date: 20250625

Claims (5)

1. The automatic health risk quotation system based on big data is characterized by comprising a multi-source heterogeneous data acquisition module (100), a dynamic health portrait construction module (200) and a risk conduction prediction module (300), wherein: The multi-source heterogeneous data acquisition module (100) is used for acquiring user behavior data in real time, wherein the user behavior data comprises sleep depth data, health behavior data and user living environment data; The dynamic health portrait construction module (200) calculates dynamic weight coefficients of different data in user behavior data by using a dynamic weight distribution method, and the dynamic health portrait construction module (200) performs space-time alignment processing on the user behavior data, unifies the data with different time granularities into daily dimension user behavior vectors, and generates weighted daily dimension user behavior vectors based on the dynamic weight coefficients; The dynamic health portrait construction module (200) calculates the risk probability value of each disease dimension through a disease risk mapping model based on the weighted daily dimension health behavior vector, aggregates the risk probability value into a real-time dynamic health risk score, and simultaneously extracts the fluctuation characteristic of the vector in a continuous preset time period to generate a health behavior entropy value; The risk conduction prediction module (300) inputs the health behavior entropy value into a time sequence prediction model, and identifies abnormal duration of the health behavior entropy value continuously lower than a preset threshold value, and outputs a personalized risk evolution path diagram based on the mapping relation between the abnormal duration and the disease development rate; The dynamic health portrait construction module (200) comprises a dynamic weight distribution unit (201), wherein the dynamic weight distribution unit (201) is used for calculating and dynamically adjusting weight coefficients of different rows of data dimensions in real time according to individual characteristics of users, data timeliness and correlation with specific disease risks; The dynamic health portrait construction module (200) comprises a risk portrait and entropy value generation unit (203), wherein the risk portrait and entropy value generation unit (203) is used for inputting weighted daily dimension user behavior vectors into a disease risk mapping model and calculating risk probability values of each preset disease dimension; The risk portrait and entropy value generating unit (203) forms a time sequence data window by collecting weighted daily dimension user behavior vectors of continuous preset time periods, calculates statistical variation indexes of all behavior features in the window, calculates average values of all feature variation indexes to obtain preliminary fluctuation measurement, inputs the preliminary fluctuation measurement into an information entropy formula, and generates a health behavior entropy value; The risk conduction prediction module (300) accesses a disease development rate knowledge base, the knowledge base stores a deterioration progress model of the disease with different dimensions, wherein the model quantifies an influence coefficient of abnormal duration of specific duration on the disease development rate; the risk conduction prediction module (300) extracts time interval data between adjacent risk upgrading nodes from the personalized risk evolution path diagram, and converts the time interval data into an accurate factor reflecting the risk acceleration degree through an accurate conversion algorithm.
2. The big data-based health risk automatic quotation system according to claim 1, wherein the dynamic health portrait construction module (200) comprises a space-time alignment and vector weighting unit (202), the space-time alignment and vector weighting unit (202) is used for unifying user behavior data collected at different frequencies and different time points to a daily dimension time frame through an aggregation technology to form a daily dimension user behavior vector, and weighting the data in the vector by applying a dynamic weight coefficient to generate a weighted daily dimension user behavior vector.
3. The automated big data based health risk quotation system of claim 1, wherein the risk conduction prediction module (300) monitors the sequence of health behavior entropy values in real time via a time series prediction model, automatically detects segments of health behavior entropy values below a preset stability threshold at consecutive time points, marks an abnormal duration and records the duration and timestamp thereof.
4. The big data based health risk automatic quotation system of claim 1, wherein the risk conduction prediction module (300) applies the refinement factor to a pricing engine and performs the following operations: Dividing the premium ladder nodes according to the predicted risk level and the predicted arrival time point, adjusting the basic premium rate by applying a fine calculation factor in the premium period corresponding to each premium ladder node, calculating the ladder premium, and generating a future premium schedule containing the time nodes and the ladder change corresponding to the new premium amount.
5. The big data based health risk automatic quotation system of claim 1, wherein the risk conduction prediction module (300) recalculates the calculation factor and the personalized risk evolution path graph upon new data input, dynamically updating the step-wise premium gradient for ensuring that it matches the user real-time risk trajectory.

Description

Health risk automatic quotation system based on big data Technical Field The invention relates to the technical field of big data, in particular to a health risk automatic quotation system based on big data. Background Currently, a fixed premium pricing mode based on a static precision meter is generally adopted in the health insurance industry, and risk assessment is seriously dependent on limited historical data (such as physical examination reports and past medical history) in the initial stage of application, and cannot incorporate daily behavior data (such as dynamic indexes of sleep, exercise, environmental exposure and the like) of a user in real time. Firstly, the bottleneck in the prior art is characterized in that the bottleneck is that the multi-source heterogeneous health big data is lack of fusion processing capacity, namely second-level physiological indexes from wearable equipment, intelligent home and mobile application, time-space dimension splitting exists between hour-level activity data and day-level environment information, the traditional system is difficult to align time granularity and quantify risk contribution weights of different data sources, and portrait distortion is caused; Secondly, the risk conduction mechanism is stiff, once the insurance fee is determined, the insurance fee is unchanged for a long time, and potential disease deterioration acceleration signals (such as regular heart rate abnormality) behind the continuous low entropy (high stability) behavior mode cannot be captured, and the predicted risk evolution path cannot be converted into preventive calculation intervention. The disjoint of the static pricing and the dynamic health track not only causes the reverse selection of high-risk people, but also causes an insurance company to miss a window period for reducing the claim settlement risk through behavioral intervention, and restricts the innovation of products and the sustainability of markets. Disclosure of Invention The invention aims to provide a health risk automatic quotation system based on big data, which solves the problems in the background technology, and the specific technology comprises the steps of realizing the dynamic weight distribution and space-time unification of multi-dimensional behavior data of a user so as to solve the problem of health portrait distortion caused by inconsistent evaluation difference of heterogeneous data value and time granularity, and fusing behavior stability analysis and risk conduction prediction so as to solve the problem that the traditional premium model cannot respond to the dynamic matching of individual real-time risk tracks. In order to achieve the above purpose, the present invention provides the following technical solutions: The health risk automatic quotation system based on big data comprises a multi-source heterogeneous data acquisition module, a dynamic health portrait construction module and a risk conduction prediction module, wherein: the multi-source heterogeneous data acquisition module is used for acquiring user behavior data in real time, wherein the user behavior data comprise sleep depth data, health behavior data and user living environment data, and a complete data base is provided for subsequent analysis. The dynamic weight distribution unit in the dynamic health portrait construction module calculates and dynamically adjusts weight coefficients of different rows of data dimensions in real time according to individual characteristics of users, data timeliness and correlation with specific disease risks, and solves the problem of individuation loss caused by static weights, for example, heart rate data weight of the users with cardiovascular disease history is higher than that of dietary habit data. The time-space alignment and vector weighting unit in the dynamic health portrait construction module unifies user behavior data of different frequencies and time points such as second level, hour level and the like to a daily dimension time frame through an aggregation technology to form a daily dimension user behavior vector, and applies a dynamic weight coefficient to weight and calculate each data point in the vector to generate a weighted daily dimension user behavior vector so as to eliminate the time-space difference of multi-source data. The risk portrait and entropy generating unit in the dynamic health portrait constructing module inputs the weighted daily dimension vector into a disease risk mapping model to calculate the risk probability value of the preset disease dimension; In addition, the risk portrait and entropy generating unit collects weighted daily dimension user behavior vectors of continuous preset time periods to form a time sequence data window, calculates statistical variation indexes of all behavior features in the window, calculates average values of all feature variation indexes to obtain preliminary fluctuation measurement, and inputs the preliminary fluctuation measu