Research On Health Insurance Sales Forecasting Based On Improved PSO Optimized ARIMA-LSTM Hybrid Model

In recent years,along with the national policy support and development of our economy,the commercial health insurance has developed vigorously.As the largest insurance company in Jiangxi Province,T Company is closely following the new development trend of the health insurance market,and its sales work faces challenges.In this paper,the improved ARIMA-PSO-LSTM combined prediction model is applied to the field of insurance.By analyzing the historical data of T Company’s sales volume of health insurance and predicting its future trend,it not only helps T company to formulate its development strategy for the next stage,but also provides a reference for future research on insurance sales forecast.Related research work and achievements of this paper are as follows:(1)Research on Integrated Auto-regressive Moving Average prediction model:This paper uses ARIMA model to forecast the time series of health insurance sales.The results show that the ARIMA model can only reflect the general trend of the whole sales volume,and can not capture the sudden change of short insurance sales volume well.Therefore,the linear ARIMA model is not effective in predicting the time series of short insurance health insurance sales.(2)Research on LSTM prediction model based on improved particle swarm optimization algorithm: This paper selects and adjusts key parameters of the neural network model,such as optimization algorithm,activation function and hidden layer number,and optimizes the super-parameters of the model through the improved and optimized particle swarm optimization algorithm.The optimized parameters were used to build the neural network and forecast the sales volume of health insurance.The prediction results showed that MAE,MSE,RMSE and MAPE values of the improved PSO-LSTM model were all smaller than the errors of the LSTM model,and each index decreased by 1.72%,0.37%,0.18% and 19.44% successively.The model optimized LSTM based on PSO improves the prediction accuracy and has better adaptability to the financial short-term insurance data with strong volatility.(3)Research on the combined model based on the ARIMA-PSO-LSTM: In this paper,the ARIMA model and the improved PSO-LSTM model were combined respectively by means of parallel weight addition and error correction,and the sales of health insurance were predicted.The prediction results showed that the MAE,MSE,RMSE and MAPE values of the error-corrected combined model were all smaller than the errors of the shunt weighted combined model.Each index decreased by 31.41%,38.69%,17.77% and 32.19%,respectively.The ARIMA-PSO-LSTM model combined with the modified prediction method is obviously superior to the parallel weighted model,and this model has the feasibility and superiority.Compared with the single model and the parallel weighted combination model,the ARIMA-PSO-LSTM model combined with the modified prediction method proposed in this paper can greatly improve the accuracy of short risk health insurance premium prediction.