Research On Customer Portrait Modeling And Renewal Prediction Of Auto Insurance

As an important research aspect of the insurance industry,auto insurance renewal is an important basis for insurance companies to formulate business strategies,and plays an important role in promoting the development of insurance companies.Therefore,it has become a research hotspot in the industry in recent years to model the portrait of auto insurance customers,explore the influencing factors of auto insurance renewal and realize fine marketing.Aiming at the problem of auto insurance renewal,this thesis conducts research from two aspects: portrait modeling and renewal prediction,analyzes auto insurance renewal from two different perspectives,and selects the renewal data set of a listed insurance company customer to carry out empirical evidence respectively,summarize the empirical results and make recommendations.In the part of customer portrait modeling,this thesis improves the traditional clustering algorithm,combines the SOM neural network in the deep learning method with the traditional K-means clustering algorithm,and proposes an improved SOM-K-means clustering algorithm,and use it to model auto insurance customer portraits to achieve more accurate customer classification.By comprehensively considering the three evaluation indicators of CH score,DBI index and silhouette coefficient,it is found that the classification effect of SOM-K-means combined clustering algorithm is better than that of single SOM algorithm and K-means algorithm.According to the final clustering results,a customer segmentation portrait is formed,the three types of customers are formed into three corresponding relationships according to the customer value,and the customer value segmentation portrait is constructed.In the part of auto insurance renewal prediction,this thesis introduces machine learning models,integrated learning models and deep learning models for insurance renewal prediction,and uses Voting algorithm and Stacking fusion algorithm to combine a single model,and compare the prediction effect of the single model and the combined model.This thesis also introduces two sampling methods,SMOTE and SMOTETomek,to balance the dataset,and compare the pros and cons of different balancing algorithms.After comparative analysis and balancing of samples,it is found that the classification effect of the model has been significantly improved.In the process of comparing and analyzing the balance algorithm,this thesis draws the conclusion that the effect of the comprehensive sampling algorithm is better than that of the oversampling.From the perspective of classification accuracy,the combined model works best,followed by the ensemble learning model,the deep learning model,and finally the machine learning model.Finally,according to the forecast results of each customer’s insurance renewal by the combined model,this thesis constructs the renewal churn portrait of auto insurance customers,and finds that the distribution of customer renewal churn and the customer value segmentation portrait form a corresponding relationship.This thesis integrates the research on auto insurance customer portrait modeling and renewal prediction,combines the customer value segmentation portrait with the renewal loss portrait,studies the internal loss distribution of various customers,puts forward the renewal loss early warning,and the targeted renewal guarantee marketing strategy.