【摘要】 针对粮食产量影响因素复杂、随机波动大等特点,对粮食产量预测问题展开研究。用混沌理论对原始样本进行相空间重构,确定最佳的嵌入维数和延迟时间。发挥粒子群算法全局搜索能力强的优点,用PSO算法优化SVM参数,避免人工选取参数的盲目性。以某省2004—2015年粮食产量预测为案例进行仿真试验,并将预测结果与灰色GM(1,1)模型进行对比。结果表明,本文所建模型对2014年、2015年粮食产量预测结果相对误差分别为-6.38%和2.07%,MAPE为4.22%,优于灰色GM(1,1)模型,具有较高的预测精度,从而验证所提方法的先进性和有效性。更多还原

【Abstract】 Regarding on the characteristics of interaction factors’complexity and big fluctuation of food yield data,method for food yield prediction is studied.Phase space of original sample is reconstructed by chaotic Theory.Optimal delay time and embedding dimension are got.PSO has ability in global searching,so it is used to optimize parameters of SVM,which can avoid blindness of choosing parameters by predictor.Taking prediction of a province’s grain yield from 2004 to 2015as an example.Prediction results are compared with GM(1,1).Simulation results show that relative error of yield food prediction for 2014 and 2015are-6.38% and2.07%.MAPEis 4.22%.Prediction accuracy of the method proposed above is higher than GM(1,1).So the advantage and validity of the method is verified.更多还原