摘要
基于传统K最近邻算法,针对农产品价格波动符合时间序列的特点,在通过计算相似度决定最近邻的时候,采用多项式函数和欧氏距离结合的方法,并用粒子群优化算法对多项式函数系数、K值的选取进行参数优化,得到改进的预测模型。实验表明,改进的预测模型的预测误差为0.281 46,传统模型的预测误差为0.371 93,预测精度提高了0.090 47,其预测稳定性强,预测精度能够达到神经网络模型的效果。
Based on the traditional K nearest neighbor algorithm, and according to the characteristics of fluctuation in agricultural product price in line with the time series, we use the methods of polynomial function and the Euclidean distance to decide the nearest neighbors by similarity, and use the particle swarm optimization algorithm to optimize the coefficient of polynomial function and K value, and finally get an improved forecasting model. Experiments show that the prediction error of the improved prediction model is 0.281 46, improved by 0.090 47 than the traditional model, whose prediction error is 0. 371 93. The prediction accuracy of the model can reach the level of neural network model, and its prediction stability is superior to the neural network model.
出处
《济南大学学报（自然科学版）》
CAS
北大核心
2014年第2期114-117,共4页
Journal of University of Jinan(Science and Technology)
基金
山东省自然科学基金(ZR2011FL016)
关键词
K最近邻算法
农产品价格
时间序列
欧氏距离
多项式函数
粒子群优化算法
K nearest neighbor algorithm
the price of agricuhural product
time series
Euclidean distance
polynomial function
particle swarm optimization algorithm
