基于粗糙集和支持向量机理论的物流需求预测研究  

Study on the Logistical Forecasting Method based on Rough Set Theory and Support Vector Machine(SVM)

作　　者：钟映竑[1] 黄鑫[1] 

机构地区：[1]广东工业大学管理学院,广东广州510520

出　　处：《工业工程》2015年第2期28-33,共6页Industrial Engineering Journal

基　　金：教育部人文社科规划资助项目(12YJA630199)

摘　　要：选择合适的预测模型来预测物流需求,对升级和优化物流产业具有重要的战略意义。常见的物流预测方法有:增长率法、移动平均法、时间序列法等,由于实际的物流预测数据常常具有多指标、非线性、小样本的特点,并且数据中存在冗余指标(噪声),导致在实际应用中,大多数预测方法的预测精度不高,难以保证有效性。针对这类物流预测问题,本文根据粗糙集属性约简中基于差别矩阵的算法,剔除冗余指标,基于约简的属性,改进了单一的SVM预测模型,并用遗传算法优化了SVM模型的输入参数,获得了较高的预测精度。本文给出了该方法的具体步骤,并用实际数据预测了广东省的货运总量,验证了该方法的有效性。Proper forecasting models are of strategic significance for upgrading and optimizing logistical in- dustry. Common forecasting methods include increasing rate method, moving average method, time series method, etc. In real applications, many forecasting methods are not so accurate as to ensure validity be- cause logistical data have such features as multi attributes （including redundant attributes）, non -linear, small sample. A method is proposed that eliminates redundant attributes for reduction based on discern- ibility matrix algorithm with rough set theory improving the SVM model and optimizing the input parameters by genetic algorithm. Detailed steps of the forecasting method are provided and validity examined via cargo data of Guangdong province.

关 键 词：物流需求预测 粗糙集属性约简 支持向量机 

分 类 号：F201[经济管理—国民经济]