Decision And Analysis Of Metallurgical Products Price Based On Artificial Neural Networks

By the economic crisis2008, China’s steel industry has been given a huge impact, which seriously affected the downstream industry-machinery industry, automobile manufacturing, construction, transportation and other heavy industries which affect China’s economic lifeline. Under the support of the country and upturn of the global economy steel industry gradually picked up, in2009. Steel industry as a pillar industry of national economy affects the development of whole national economy and many downstream industries. In order to forecast the metallurgical products price trend accurately, this paper introduces intelligent information processing technology and use artificial neural networks and genetic algorithms to build a neural network model, which can help the iron and steel enterprises grasp market trends early and to make the right business decisions as soon as possible.First, this paper analyzes the factors impact the metallurgical products price, and extracted the neural network forecast model’s input parameters from five major influencing our country’s metallurgical products price aspects, the macro-economic, fixed asset investment, supply and demand raw material and international market. They are the gross domestic product, consumption spending, fixed asset investment, crude steel production, steel export volume, the domestic iron ore production, iron ore import volume, iron ore price change rate of international agreements.Then this paper chose BP (Back-Propagation) neural network to establish a forecast model. The number of layers and parameters of the model were determined by experiments. Then Levenberg-Marquardt algorithm was used to improve traditional BP training algorithm, which reduced the training time and improved the accuracy of convergence. After that, genetic algorithm is used for a combination with neural networks to optimize BP neural network initial weights and threshold value. Then a price forecast model was established for metallurgical products.Finally, this paper use program in MATLAB to simulate, train, validate the model established and forecast the metallurgical products price. After several experiments, the results proved that the model established in this paper has a fast training speed and high forecast accuracy, which can used in metallurgical products price in China for highly accurate predictions.