The Application Of Neural Network Technology In The Coal Chemical Industry Product Price Forecasts

In recent years, the voice on using China's own coal resources to replace imported oil has been louder and louder, coal is expected to be a new energy of our country, the increase of oil price as well as the demand growth of the coal and coal chemicals will bring new opportunities and challenges to the development of the coal industry. Writing China's coal market analysis report has been very hot nowadays. The most importance thing is concerning about the price forecasts in the report. Anyway, in market economy, how to track and predict the market trend of coal chemicals scientifically?Using industry background, this paper attempts to building system about price forecast. The aim is to be data mining from searching all history data about the coal chemicals. The work of the paper is as follows:(1) Non-linear characters of the coal chemicals price time series can be determined according to BDS methods. Besides, the method to combine neural networks model and time series model is presented.(2) Using factor analysis approach, the training and testing sample of the neural network can be identified. Then, pay more attention on doing the data preprocess, typically on combining three variables into one new variable. About identifying of Outliers, such a method counts the difference between the former and certain behind points of all factors.(3) Defining BP neural network structure according to the state-space reconstruction and the L-M algorithm. By training, testing and predicting, the result indicates for sure that the model is efficient that it can catch highest point and lowest point of the pricing data. It also turns out that the neural networks of multivariate time series model is better than the neural networks of univariate time series model and the index smoothing model.