Application Of Data Mining In Forecasting The Trend Of Sea Product Price Index

With the development of economy,Seafood trade of China is changing towards internationalization and entering the stage of rapid development,which has led to the upgrading of the aquatic industry structure and enhanced competitiveness.As an important parameter to measure the market operation,the price index of the product is closely related to the supply and demand of the market,the economic interests of the enterprise and the mode of production.Therefore,forecasting seafood price index is not only beneficial for enterprises to constantly adjust production and sales mode,formulate reasonable marketing strategy,but also conducive to macro-control of the market,effective allocation of resources,and then promote continuous and stable operation of seafood market.In this dissertation,taking three crab price index and the six factors from October2012 to September 2017 as an example,we analyze the correlation between seafood price index factors,then use grey prediction,BP neural network and wavelet neural network three methods for the next year's price index forecast.Firstly,establish the gray GM(1,1)model and take directly the price index as the original data.After the matrix transform,price index will be taken as a solution to first-order differential equation.Determining the development of grey number and endogenous control of grey number,the equations are given concrete form.Time parameter shift after 12 units can come next year's price index forecast value.Secondly,a three-layer BP neural network model is established.After normalizing the price index and influencing factors,the number of the best hidden layer nodes is selected according to the empirical formula,and a circular code is set to accomplish this goal.According to the actual needs,we determine the weights and thresholds of the network,adjust the transfer function of input layer,hidden layer and output layer,train the optimal network model,complete the construction of BP neural network,and predict the next year's price index.Finally,the wavelet neural network model is similar to the BP neural network model,and the wavelet neural network is also set to three-layer model.According to the principleof wavelet transform,we determine the wavelet expansion factor,translation factor,the connection weight of the network,load the normalized raw data,and train and modify the network parameters.On the basis of the prediction results of the three forecasting models,the effect of the model is compared and analyzed.It is found that the prediction accuracy of wavelet neural network is superior to the other two models in accuracy and credibility,which has practical value and significance for future market prediction analysis.