Research On Agricultural Product Price Forecasting Model Based On Deep Learning

With the development of world economy,China's foreign economic and trade exchanges are also increasing.Since the accession to WTO,China's economic development can be said to be the parallel opportunity and development.The consumption level of domestic masses increases,and economic globalization affects all kinds of product market.As the consumption necessity,agricultural products greatly occupy the influence,but a large number of agricultural products market is facing the challenges accordingly.The agricultural prices is not stable and the market price volatility is high,so that the price adjustment for agricultural products is becoming a hot spot of the market research,and stabilizing agricultural products market price draws more and more attention of the national government.In many agricultural products,vegetables and pork are indispensable food in people's daily life.Its price fluctuations will bring adverse effect on agricultural market's production operators and consumers.Market price prediction research on agricultural products has an important significance on stabilizing agricultural products market order,ensuring citizen's basic livelihood stability and steadily improving farmer's income standard.Deep Learning as a kind of new multilayer neural network learning algorithm,compared with the previous algorithm,can make up for the defects of traditional algorithm that easy to get into local minimum.It can well realize the representation of complex high-dimensional functions such as high variable function,and reduce the complexity of the calculation.In view of this,in this thesis,by introducing deep learning algorithm,and its self-learning feature,we determined the network optimal structure of deep learning based on the agricultural products price data from 2005 to2012,and the data of the 2013 and 2014 are used as test samples to achieve a more accurate prediction of agricultural products prices.This thesis mainly studies from the following aspects:First,in this thesis,the basic theory of Deep Learning is introduced in detail.The basic concepts,structure and derivation of several common deep learning model,such as Boltzmann Machine,Restricted Boltzmann Machine,Convolutional Neural Network and Deep Belief Network.Second,based on the analysis of the formation of agricultural products and the mechanism of volatility,so as to determine the main factors affecting the price of agricultural products.By collecting the data of vegetable prices,pork prices and theirinfluencing factors,we normalize these data,distributing them in the range of [0,1],then carried out the principal component analysis to extract the number of the data components,so as to eliminate the data Collinearity.Third,taking the vegetable prices,pork prices and its principal component from2005 to 2012 as the correction set,this thesis determined the number of the optimal hidden layer and the number of nodes contained in the hidden layer of Deep Belief Network by the experimental method.and took the data in 2013 and 2014 as the test set.The result shows that the prediction accuracy of the deep learning network model is more accurate than BP neural network and wavelet neural network prediction model.Studies have shown that the deep learning network model has a very good prospect in agricultural products price forecast and other commodity price forecasting.And also the establishment and research of agricultural product price forecasting model can provide a reference for other product market.This thesis mainly committed to the establishment of the price forecast model of agricultural products and the studying of the more excellent models to forecast the price.This research will conducive to the steady and efficient development of China's agricultural products market,which is of great significance to China's economic and social development and national long-term stability.