Research And Building Of Loss Ratio Prediction Model Based On Time Series

With the continuous increase of competitors in PICC Ningxia Branch and the worsen competition situation,the strengthen of data analysis and forecasting is urgently needed.Fluctuation of loss ratio determines the insurance companies operating level.The establishment of the loss ratio prediction model can enhance planning,reduce blindness and enhance profitability.The main work of research and building loss ratio prediction model based on time series includes:1.Prediction research and analysis.By comparing the advantages and disadvantages of regression and time series predicting model,combined with the actual situation of the loss ratio sequence.The time series predicting model were mainly studied.The steps of specific time series predicting models including:data acquisition,steady judgment,identification,estimation,checking,correction and prediction.Providing a theoretical support by the analysis methods of deterministic time series and random sequence were established.2.Building the loss ratio prediction model by calculation method.Prediction of the loss ratio can be separately predict comprehensive indemnity and earned premiums,and the loss ratio is divided by the comprehensive indemnity and earned premiums.In the prediction of comprehensive indemnity and premiums earned,using different model identification method,like exponential smoothing method and different parameters ARIMA model,then determine the Holt-Winters multiplicative model of the time series fit better.So do the model estimation,model inspection and model forecast by Holt-winters multiplicative model,separately for comprehensive indemnity and earned premiums.Finally,calculating the loss ratio fitted values average relative error is 0.0522 from 2013 to 2014,predicted values average relative error is 0.0510 from January to April,2015.3.Building the loss ratio prediction model by loss ratio value method.To predict the loss ratio can be obtained through the analysis of historical loss ratio sequence.First,judge the stability and periodicity of the loss ratio sequence,then,analyze the sequence of the loss ratio is ARIMA model.But,the model test showed significant model is not high,so Introduce the independent variables(comprehensive indemnity and earned premiums)to re-identification,re-estimation,re-testing of the model,and obtained the ARIMA(1,1,0)prediction model with the independent variable.Using this model,calculating the loss ratio fitted values average relative error is 0.0139 from 2013 to 2014,predicted values average relative error is 0.0434 from January to April,2015.Based on the research above,compared with the calculation method and the value method of building loss ratio prediction.The loss ratio value method has less error between fitted values and predicted values,and also has the better model fitting degree.This model is able to meet the actual demand and provide the reference basis for the loss ratio control of PICC Ningxia Branch.