Demand Prediction Of Agricultural Insurance Based On GA-LSTM Model

On February 21,2021,the CPC Central Committee and The State Council issued the No.1 central document guiding the work related to agriculture,rural areas and farmers,Opinions on Comprehensively Promoting Rural Revitalization and Accelerating Agricultural and Rural Modernization.The document pointed out that rural areas must be revitalized if the nation is to be revitalized,and the most important thing to realize rural revitalization is to solve problems related to agriculture,rural areas and farmers.Agricultural insurance calls again "agricultural insurance",it is to be engaged in agricultural production process only for agricultural producer,to suffer natural disaster,accidental epidemic disease,disease to wait for a kind of economic loss that insurance accident place causes to offer safeguard insurance.The study of agricultural insurance plays an important role in understanding the solution of the problems concerning agriculture,rural areas and farmers.The development of its business is conducive to dispersing the risks of agricultural production,stabilizing farmers’ income and promoting rural development.At the same time,the study of its demand is helpful for insurance companies and the government to make macro-control and policy to grasp the input capital and subsidies.In order to accurately grasp the demand situation of agricultural insurance in Jiangxi Province,the effective way is to select appropriate evaluation indicators and models to analyze and forecast the demand of agricultural insurance in Jiangxi Province.First of all,in this thesis,the development status quo of jiangxi agricultural insurance demand,related conclusions:(1)in the three stages of its development,jiangxi agricultural insurance income which agricultural insurance demand there is an upward trend in these years,the policy compensation demand increased year by year,planted demand diversification,subsidies also constantly improve,constantly improve agricultural insurance system mechanism.(2)There are some problems in the demand and supply of agricultural insurance in Jiangxi Province,including the risk awareness of farmers in the demand level fails to keep up with the changes of the environment,and the gap between the ideal payment and the reality is large;At the supply level,too high market concentration will affect the market vitality,thus reducing product innovation ability,insurance companies agricultural insurance business losses,agricultural insurance product coverage is not enough.Secondly,based on the agricultural insurance data of Jiangxi Province from 2001 to 2019,this thesis uses deep learning method in agricultural insurance demand prediction,and proposes a LSTM neural network model optimized by genetic algorithm.At the same time,through a general overview of the current situation of the development of agricultural insurance demand in Jiangxi province,and the characteristics of the agricultural insurance itself to comb,and on the basis of the real agricultural insurance demand forecasting,this article from the efficiency of agricultural insurance,government support,agricultural production and farmers’ own factors such as dimension of four constructs the index system of agricultural insurance demand forecasting.On the basis of 16 explanatory variables in four dimensions,ga-LSTM model was constructed,and its prediction results were compared with those of single LSTM model and traditional machine learning model.The study concluded that:(1)When the 16 factors included in the agricultural insurance demand evaluation index system summarized in this thesis are used as the input layer of the neural network,the LSTM neural network model can be used to better predict the agricultural insurance income.(2)Genetic algorithm optimization of hyperparameters can effectively improve the prediction ability of LSTM neural network,root mean square error RMSE is 0.2968,mean error MAE is 0.2215,goodness of fit R^2 is 0.9884.(3)The prediction accuracy based on GA-LSTM neural network model is more effective than the single LSTM model and the traditional machine learning prediction method.