Prediction Of Individual Payments Of Non-life Insurance Claim Reserving Based On SVM

Insurance is an effective means to combat risk under of market economy.Non-life insurance is an important part of insurance company's business.As a major liability of insurance company's operation,the claim reserving assessment plays a vital role in insurance practice.Aiming at the individual claim reserving for non-life insurance claims,this paper implemented and evaluated a machine learning model integrated with established Support Vector Machine(SVM).This work also explores the application prospect of machine learning method in the assessment of non-life insurance reserves.Focusing on the assessment of non-life individual claim reserving,background knowledge of actuarial and machine learning related to this study was introduced.Based on the current progress,a SVM integrated model was put forward to predict individual claim reserving.Some simulated data was utilized to analyze the model empirically.In the aspect of actuarial knowledge,some basics and commonly-accepted classical methods of non-life insurance claim reserving were introduced.In the aspect of machine learning,this paper briefly summarized some concepts and methods involved in this study,and Easy-Ensemble method was utilized to generate datasets from the extremely unbalanced source data.According to previous research,the model for predicting number of claims for individual non-life insurance was constructed by SVM,which was utilized for its completeness in theory.A simulated dataset was generated for empirical analysis.Based on the robustness requirement in insurance practice,confusion matrix and the recall were evaluated for the machine learning model.The experimental results showed that this machine learning model of non-life insurance individual claim reserving can meet the requirement of sufficiency and accuracy of reserve provision in well-informed forecast years.