Application Research Of Health Insurance Purchase Prediction Model Based On GA-CatBoost

With the rapid development of the Internet and artificial intelligence,information acquisition methods are more diverse and public participation is increasingly high.Every move of people has become an important product of the era of big data.China’s insurance industry is undergoing rapid development,among which health insurance is one of the most concerned in recent years.During the 13 th five-year Plan period,the insurance industry of our country,as a "stabilizer" and booster,has provided reliable risk transfer and insurance for the healthy development of our national economy.According to the 14 th Five-Year Plan,in the next five years and beyond,China will focus on developing new structural potential in line with the medium-speed growth period,and the insurance industry faces more challenges and opportunities.At present,China’s insurance industry urgently needs to optimize traditional marketing methods and advance digital marketing.At the same time,machine learning,as the mainstream method to solve many artificial intelligence problems,is widely used in various fields.In this context,this paper mainly studies the application of machine learning algorithms in the insurance industry,such as user purchase behavior prediction and influencing factors analysis.Data related to users of a company’s health insurance products were collected.Data cleaning and preprocessing were carried out on 39 variables such as personal information,family information and regional information.The Boruta wrapper combined with Shap-value was used to screen out important features,and the features that might affect users’ insurance purchase were extracted for analysis and modeling.And try to use genetic algorithm to optimize Cat Boost algorithm,obtain the insurance purchase prediction model based on GA-Catboost.It is compared with the model constructed by random Forest,XGBoost and light GBM.The results show that GA-Cat Boost algorithm has obvious advantages in the application of health insurance purchase prediction.The algorithm is compared with the unoptimized Cat Boost model,and the effectiveness of the optimization method is verified.By comparing the feature set before screening and the feature set after screening,the feature screening method is proved to be feasible.Finally,with the help of Shap model,the important factors affecting consumers’ purchase of insurance and their mechanism of action on users’ purchase behavior decisions are analyzed.This application research can help the insurance industry to improve business efficiency,identify high-quality customers,reduce the resistance of business promotion,and provide reasonable basis for insurance companies to formulate precise marketing strategies,so as to ensure the healthy development of the insurance industry.