Price Forecasting Of Agricultural Products Based On PSO And BPNN-LSTM

Agriculture is the basic industry in China.The price of agricultural products is closely related to every citizen's daily life.The price fluctuation of agricultural products is an important reason affecting the stability of the agricultural product market.Predicting the price of agricultural products will be conducive to the normal operation of the national economy.In China's agricultural field,due to the wide variety of agricultural products and the characteristics of large amount of data information and many types of data,it is very difficult to predict the price trend of agricultural products in the future by using the price information of agricultural products.At this stage,there is an urgent need for a method to predict the price of agricultural products in the future through the obtained price information of agricultural products.The specific work of this paper is as follows:(1)Research on particle swarm optimization algorithm for neural network parameter optimization.The price of agricultural products has the characteristics of large fluctuation and nonlinearity.When using neural network to predict the price of agricultural products,because the neural network uses gradient descent algorithm to update the weight,it is easy to lead to the problem of local minimum.Particle swarm optimization algorithm has good global optimization ability.The parameter optimization of neural network based on particle swarm optimization can make the neural network avoid the local extreme value phenomenon during training,so as to improve the effect of agricultural product price prediction using neural network.However,particle swarm optimization has some disadvantages,such as slow optimization speed and local convergence.In order to enhance the optimization ability and improve the prediction accuracy,a particle swarm optimization algorithm(TPSO)for neural network parameter optimization is proposed in this paper.On the one hand,the algorithm introduces the domain collision mechanism,sets the collision operator for particles and adaptively changes the domain,and controls the diversity of the population in different stages in the whole search process to prevent the premature aggregation of particles,so as to improve its exploration ability;On the other hand,by introducing a new particle interaction mechanism,particles have better local search ability in the iterative process of the algorithm.In the experiment,several benchmark functions are used to prove the efficiency and robustness of TPSO algorithm.(2)Research on agricultural product price prediction model based on deep learning theory.Based on the characteristics that TPSO algorithm is simple and easy to implement and does not have many parameter adjustments,BP,LSTM and other neural networks have good results in nonlinear time series prediction.Combine the three and give full play to their advantages.Aiming at the problems of large random fluctuation and over limit prediction error in the prediction of existing agricultural product price prediction models,An agricultural product price forecasting model(TPSO-BPNN-LSTM)based on TPSO optimized forward feedback neural network(BPNN)and long-term and short-term memory network(LSTM)is proposed.In this method,TPSO is used to optimize the parameters of BPNN and LSTM.Aiming at the advantages of strong prediction stability of optimized BPNN and high prediction accuracy of optimized LSTM,the two optimized neural networks are effectively combined to establish the best prediction model for the application of agricultural product price prediction.(3)The obtained price data of Chinese cabbage,cucumber and leek are used to evaluate the established prediction model.The experimental results show that the prediction accuracy of BPNN optimized by TPSO is improved,the prediction accuracy and calculation efficiency of LSTM optimized by TPSO are improved,and the agricultural product price model TPSOBPNN-LSTM constructed after the comprehensive use of the two optimized neural networks has high prediction accuracy It has the characteristics of strong stability and high calculation efficiency.