Research On Fruit And Vegetable Price Forecasting Based On Data Mining

Market prices of agricultural products are closely linked to people's lives.In order to ensure a happier life for the people,the government should actively respond to the call of the nation as a scientific and technological power,combine emerging science and technology with agriculture,effectively improve the ability of market monitoring and management,and vigorously promote the modernization of agriculture and rural areas.Since the beginning of the 21 st century,intelligent electronic technology has developed strongly and mobile electronic devices have become increasingly popular.Everyone will produce a large amount of data every day.However,not all of the data are of realistic significance,and excessive and complicated data will disturb judgment.In order to search redundant data,data mining technology arises at the historic moment.What data mining technology is good at is to summarize messy data,and find out useful information according to its rules,so that it can continue to serve human beings.This dissertation mainly used neural network to forecast the price of apple and potato in Jiangsu Province.Jiangsu Province is located in the temperate zone and subtropical plain,and the geographical conditions such as landform and climate are suitable for the development of fruit and vegetable planting industry.Firstly,this dissertation understood the specific market situation of apples and potatoes in Jiangsu Province,grasps the general law of product price fluctuation,and analyzed the influencing factors that lead to price fluctuation.At the same time,due to the large number of data samples needed for model training by using neural network,the network crawler technology was used to reduce errors in the process of manual collection when collecting data.Secondly,the principal component analysis method was used in the preprocessing analysis of the data.In order to reduced the interaction between the influencing factors,the independent principal components were analyzed to reduce the computational workload,and the principal component analysis was used to reduce the dimensionality.Finally,according to the characteristics of neural network model,this dissertation selected BP neural network model,RBF neural network model and NARX neural network model based on Levenberg-Marquardt algorithm to carry out price prediction experiment.After all data preprocessing is completed,the data from 2009 to 2019 are substituted into the experiment,and the BP,RBF and NARX neural models are trained by adjusting the transfer function and setting various parameters.Then the influence factor of 2019 is substituted into the trained model to obtain the prediction data of 2019.According to the comparison between the predicted value and the actual value,it is found that the NARX neural network model was suitable for predicting the price of apple with greater price fluctuation,while the BP neural network model based on L-M algorithm has smaller error in the potato price prediction experiment,which provides theoretical support for the government departments to implement macro-control policies in the future.