Research On Apple Sales Forecast Method Based On Deep Learning

In recent years,with the development of China's economy and society and the implementation of the strategy of rural revitalization,agricultural development has also entered a new stage of transformation from traditional agriculture to modern agriculture.The production of all kinds of agricultural products has also accelerated the transformation from traditional small-scale agricultural production to socialized large-scale production.Especially with the continuous improvement of scientific and technological modernization and innovation capabilities of the apple industry,apple sales prices and market trends have also changed.This paper takes the current apple market sales price as the core,and systematically analyses its price change and fluctuation law.By analyzing and predicting the market price trend of the apple industry in the future,this paper provides technical reference for the new agricultural modernization reform in China.The main work of this paper includes:1.Research methods for agricultural product price forecasting.On the basis of the commonly used price prediction models,we compared the advantages and disadvantages of all kinds of prediction models,then we developed a new model with higher accuracy and stability.We found that in the early study of the price forecasting,the difference autoregressive integrated moving average model(ARIMA)is used to the short-term prediction of the apple market sales price.But as the amount data increase,the forecasting accuracy decreases.In the later study,researchers proposed recurrent neural network(RNN),long-term and short-term memory artificial neural network(LSTM),and the attention mechanism combined long-term and short-term memory network etc.These models are able to achieve short-term prediction of apple market price on the time series.But all these models have disadvantages of time consuming and low accuracy.In this paper,a prediction model which combines the Seq2 Seq model based on the attention mechanism(Attention)with the long-term and short-term memory(LSTM)artificial neural network model was proposed to the short-term prediction of the market price of apple.By integrated learning,this model combines the network,LSTM and Seq2 Seq together effectively.Thus it can greatly improve the accuracy and stability.2.Analysis of impact factors in apple market price forecast.In the actual research,the price fluctuation of apple market not only depends on its own influencing factors,but also depends on the influencing factors of external environment.Based on the previous research results,we take the price index of Luochuan Fuji apple with paper bag in Shaanxi province as the research object,and use the multiple regression analysis to get the following influencing factors of apple market price: historical apple price,temperature,rainfall,dew point and frost point,seasonal alternative fruit(strawberry)and consumer price index(CPI).3.The Apple market price forecast and analysis.The influencing factors of apple price fluctuation is studied by the time series analysis method.In this paper,based on the deep learning,the establishment of Tensorflow framework and the integrating learning to connect the forecast models LSTM and Seq2 Seq based on the attention mechanism,a new apple market price prediction model was constructed.Besides,in order to validate the experimental accuracy,we establish the time of convolution network prediction model(TCN),which is based on the improved LSTM and the Seq2 Seq,to forecast the fitting degree prices of apple's daily price,weekly price and monthly price of Luochuan Fuji apple with paper bag in Shaanxi province.By the comparison experiment,the new model used in this paper is better than the three comparative prediction model in apple market price prediction.And the experimental results show that the errors of mean square and root-mean-square for the new model are much smaller than those of the three comparative prediction model.It also provides new ideas and methods for how to predict more quickly and effectively the short-term Apple market price under a large amount of datum.An Apple sales price forecast system was designed and developed,which contains market quotation,weather information,price forecast,pests and diseases and other information query function.Users can intuitively find the information they need.At the same time,the system can also provide some help for the related industry.