Shandong Agricultural Product Price Forecasting Model Based On KNN-BPNN Algorithm

The information revolution has brought more efficient production, storage andconsumption of date, as well as geometric explosion of data growth, and thus it was foundthat the use of the existing fragmented data caused a huge waste, people began to reach aconsensus for miming the data systematic, so generated data mining technology and broughtgreat social value. In agriculture of the whole industry chain, on the one hand with thedevelopment of agricultural information on agricultural production, the processing,distribution, marketing and other aspects of agricultural have a lot of quality data is stored, onthe other hand, the phenomenon of “the high price of vegetables is detrimental to the publicwhile the low price does harm to the peasants” has frequently occurred, indicating theutilization of agricultural data is still low. So using data mining techniques to analysis thesedata and found the valuable model for solving the question and other social phenomena, topromote agricultural production, maintaining social stability has important value andsignificance.This paper is the research of Shandong agricultural product price forecasting modelbased on data mining technology, target using data mining algorithms developed Shandongagricultural product price forecasting model which has a high prediction accuracy rate, astrong predictor of stability and low computational complexity. In this paper, based on theclassical nearest neighbor algorithm, we use the methods of polynomial function and theEuclidean distance to decide the nearest neighbors by similarity, and use the particle swarmoptimization algorithm to optimize the coefficient of polynomial function and k value, andfinally get an improved nearest neighbor algorithm. Use the particle swarm optimizationalgorithm to optimize the parameters of the feed-forward neural network which is widely usedin a variety of forecasting models. The improved nearest neighbor algorithm has predictedstrong stability characteristics and the feed forward neural network has high predictionaccuracy, therefore, an effective combination of these two algorithms used to establish theprediction model can get the best results.At last, we use the Chinese cabbage price data that get from Vegetable Wholesale Marketof Qingdao city nearly three years to assess the improved algorithm and the forecasting model. Demonstrated that the improved nearest neighbor algorithm improves the prediction accuracy,the improved feed forward neural network algorithm improves the prediction accuracy andreduces the computational complexity. After finally proved the validity and feasibility whichcombine the two algorithms on the accuracy, stability and efficiency.