Bulk Commodity Price Forecasts Based On Genetic-BP Neural Network

In recent years,the iron ore and soybean prices rose sharply and this phenomenon caused a loss on China's economy.It needs immediate concern that Chinese pricing power of the international commodity has always been at a disadvantage.Moreover,there are obvious shortages on the forecasting of international commodity's prices.If we can correctly analyze the characteristics and operation rules fluctuations in international commodity prices through effective methods,and then effectively forecast the commodity's prices,which is of a very important significance to the country's economic policy formulation and enterprise production management decisions.But factors volatility of commodity prices suffered a lot,such as policy,market,social,etc.Furthermore because of the internal structure of its complex pricing system,the traditional price forecasting tool would hardly meet the needs.In this paper,we analysis commodities price predicting in advanced scientific research and theory,study and discuss these conclusions,get learned the factors which affect commodity prices float,summarize the theory premise of forecasting price,at the same time point out the problems in price forecasting.After considering the characteristics and advantages of the genetic algorithm and BP neural network in different areas,this paper proposes a excellent combination forecasting model.This compound algorithm not only can fully exert the global search ability of genetic algorithm,but also maintain the advantages of keeping the search speed and searching the zone of ignorance;at the same time the advantages of BP neural network topology could be fully reflected,which means it further improve the search speed and precision on the good performance of genetic algorithm.The main purpose of the article is based on genetic algorithm BP neural network implementation techniques and its application in the field of commodity price forecasts.Firstly,on fusion process of Genetic Algorithm and BP Neural Network Fusion,encoding and decoding of genetic,design of fitness function and genetic operator design has been improved and innovated.Then determine the reasonable and effective input and output variables,using fusion of genetic algorithms and neural networks to determine the optimal number of hidden nodes and the network weights for commodity prices simulate forecast.Paper selected from 2009 to 2014 of the CRB spot price index which be used to verify the predictions,and the predictions will be compared based on the predicted results of genetic algorithm BP network model with single prediction model,experimental analysis show that,for bulk commodity price forecasts,the model not only has its effectiveness,feasibility and reliability,but also has good prospects for development and application value as well.