Research and Implementation of the Prediction Methods of the Vehicle Insurance Renewal Rate Based on Model Fusion

摘要

This essay is based on an insurance company policy data and as the original data to make training test via Spss, Matlab, Python and three algorithms of XGBoost, LightGBM and BI-LSTM combining the big data, neural web and other relevant knowledge and technologies. And then, we can speculate the various customers of renewal insurance rate via the relatively and robust mathematical models which are made F1 value as the reference to adjust the weight and linear fusion. Simultaneously, we obtained the accuracy is about 0.88 under many data test verifications stably. And this satisfied simulation result can indicate this way is appropriate.