The Research on Agricultural Product Price Forecasting Service Based on Combination Model

摘要

Under the market economy, the moderate fluctuations in the price were normal reaction of the market mechanism. However, excessive fluctuations in the price would also bring adverse effects to corresponding industry. The price change of agricultural products was the great event vital to national well-being and the people's livelihood. Since the price of agricultural products was affected by many factors, it was difficult to predict the price of agricultural products timely and accurately. The emergence of cloud computing, accompanied by massive storage space, huge computing capacity and low cost, provided a new direction for data mining to deal with big data. To realize the short-term agricultural product price forecasting, took service oriented forecasting as the guidance, this paper constructed a agricultural product price forecasting service system based on the DOA architecture. The system could conduct the analysis and prediction for the market price through the combined forecasting method of wavelet transform and BP neural network. Based on the 72 sets of monthly price data from January 2013 to December 2018 of the spinach, cabbage, tomato, pepper and potato, this paper studied the changing trend of agricultural products price. Firstly, in this paper, the price was conducted the db5 three scale decomposition using wavelet transform. Then, the shredded trend part and detail part were predicted by BP neural network model respectively. Finally, the prediction results of each component were combined and reconstructed. On the basis of this combined forecasting method, the agricultural product price forecasting service system with functions of price data service and price forecasting service was designed based on DOA architecture. To verify the prediction accuracy to the price of the 5 kinds of vegetables, the prediction accuracy index was used to evaluate and analyze the price prediction results of the system. The results showed that the minimum mean absolute error of the system was 0.083 yuan/kg, the minimum mean percentage error was 3.95%, the minimum mean square error was 0.102 and the agricultural product price forecasting service system combining wavelet transform and BP neural network had better agricultural product price forecasting performance. At the same time, this also meant that the universality of this system could adapt to the price forecasting of a variety of vegetables and provide technical support for producers, operators, consumers and relevant government departments.