基于参数化三角范数的铁路货运量选择集成预测

Selective Ensemble Forecasting of Railway Freight Volume Based on Parametric Triangle Norm

针对铁路货运量预测中预测方法单一、准确度不高、泛化能力弱问题,基于参数化模糊逻辑理论,结合前序法选择策略,提出了一种新的基于Yager三角范数的选择性集成学习模型,并应用于铁路货运量预测.采用5种常用的单预测模型作为候选基学习机模型,以误差率作为评价指标,采用前序选择策略选定2种基学习机进行集成预测;以遗传算法和最小二乘法确定集成模型的参数,实现铁路货运量预测基学习机的最优组合.试验结果显示,对比单预测模型、最优组合预测模型和均方误差导数预测模型,新提出的选择集成模型取得了最低的误差率,表明其在铁路货运量预测中能够有效提高预测精度.

Focused on issues of railway freight volume forecasting that the forecasting method is singleness,theaccuracy is not high enough and the generalization ability is poor,a new selective ensemble learning model based onYager triangular norm is proposed,and is applied to the forecasting of railway freight volume.Its candidate basemodels including five kinds of single forecasting model commonly used are adopted,using deviation rate as evaluationindex,and two kinds of base learning machines selected by forward sequential selection policy are used to forecastintegrally;The genetic algorithm and least square method are used to confirm the parameters of ensemble model,theoptimal combination of the railway freight volume forecasting is achieved.Experiment results show that the newlyproposed selective ensemble model has achieved the lowest deviation rate,compared to single forecasting model,optimal combination forecasting model and mean square error derivative forecasting model.This indicates that itcan effectively improve the forecasting accuracy in the railway freight volume forecasting.