The Modeling And Improvement Of BP Neural Network Prediction For The Price Of Pork Which Combined With Gray Theory

In recent years, along with data mining technology continues to evolve and progress, has brought the people's production and life tremendous economic benefits and convenience. To study and improve the data mining technology also attracted more and more people's enthusiasm and concern. A large number of scholars and research institutions are researching on the data mining technology improvement. With the people's continuous efforts, some data mining technology has become increasingly mature. For example, neural network model, decision tree model, evolutionary computation models and so on.Now the data mining modals are continuing to developing and progressing towards a broader applicability and a more specialized application in both directions. In view of data mining model corrective method, the people also propose various plans. Unifies various data statistics technology or the data technology utilizes in the data mining model, possibly causes the original certain technical ability of data mining model to obtain the big improvement and promotion.Neural network model is one of the most commonly and widely used data mining models. The most prominent characteristic of this model is the automatic ability to learn. Because it uses a relatively simple way to solve the big complexity problem, neural network model can solve complex problems easily, which have hundreds of parameters. Today neural network model is mainly applied on the classification, regression, forecasting, pattern recognition and other fields, it has made a huge success.BP(Back Propagation) neural network (error back-propagation algorithm based on artificial neural network) is a feed-forward type of neural network learning algorithm, which made by team of scientists led by Rumelhart and McCelland in 1986, it is a kind of error back-propagation algorithm according to the training of a multi-layer feed forward neural network. BP neural network is currently one of the most widely used neural network model.Gray theory is a theory of applied mathematics which researches the information who has both clear part and vague part and uncertain phenomenon. Ever since 1982 it made by Deng Julong who was the Professor in China Huazhong University of Science and Technology, who caused a great response around the world. experts and scholars have given this theory a high appraisal. Since the objective world, the gray system (information both has certainty part and uncertainty part) exist in large numbers, so the gray theory has become a sword for researching in this area. Today, various methods based on gray theory has been widely applied to different areas of scientific research among different disciplines, it has accessed to a lot of gratifying results.In this paper, First of all, I detailedly introduced the concept of data mining, I compared the data mining technology with traditional data statistical techniques, OLAP technology, data warehouse technology and analyzed the difference between these technologies and applications etc. Then I introduced and researched the neural network, gray theory and BP neural network techniques, that would be as the necessary preparations knowledge for building up the data mining model.The Core part of this article uses given data to establish a set of complete data mining model for market price forecast. The process covered by the definition of data mining issue, the creation of data mining database, analyzing data, preparing data, building data mining models, the evaluation of model and other steps.In this paper first I defined the data mining problem to establish a market price for pig forecast data mining model, and based on this issue I decided to adopt a neural network model to build model.The process of building data mining database also includes data collection, data description, data selection, data assessment and data cleansing, data consolidation and integration, building meta data and developing data mining database, maintaining data mining database and so on. In this paper, I have carried out a brief description on the various steps.The process of analyzing the data, this paper adopts multiple charts to intuitively show a variety of data and get ready for data analysis in next steps later.In the preparation processes of data, in order to fit for the requirements of network data input, this paper has made the normalized processing on a variety of data. this paper first I discussed GM(1,1) prediction model of gray theory, and then presented an improved GM(1,1) prediction model. I used this improved GM(1,1) algorithm to deal with the problem of Missing values. Relative to the original GM (1,1) algorithm, the improved algorithm significantly improves the accuracy of missing value fill.In the process of building data mining models, this paper used a BP neural network algorithm based on gradient descent algorithm, and then I combined the improved GM (1,1) prediction model and the BP neural network model to carry out the Improvement of the learning algorithm of neural network.At the end of this article I made a summary and prospects.