MRA-RBF神经网络组合算法的煤矿材料成本预测  

Prediction of coal mine material cost based on MRA-RBF coupling algorithm

作　　者：王乐[1] 任海芝[1] 

机构地区：[1]辽宁工程技术大学工商管理学院,辽宁葫芦岛125105

出　　处：《辽宁工程技术大学学报（自然科学版）》2017年第5期548-553,共6页Journal of Liaoning Technical University (Natural Science)

基　　金：国家科技支撑计划(2013BAH12F01);国家自然科学基金(51374121)

摘　　要：为解决煤矿材料成本预测存在的问题.将多元回归模型和RBF神经网络相结合,建立了煤矿材料成本预测的MRA-RBF耦合模型.从自然因素、技术因素、管理因素等方面选取8个变量建立煤矿成本预测指标体系.对实际煤矿材料成本进行预测分析.结果表明:MRA-RBF耦合模型预测最大误差为10.795 145 2%,平均误差为5.459 71%,最小误差仅为0.344 581 7%,预测效果较好,预测精度与单一MRA模型及RBF神经网络相比有了较大提高.验证了所提出模型的科学性、准确性,说明将线性拟合算法(MRA)和非线性拟合算法(RBF)结合起来用于煤矿材料成本预测是一种较为优越的算法,为煤矿材料成本预测及控制提供一种新的方法.Coal mine material cost forecast as the basis of budgeting is a very important link of coal material cost management. In view of the present coal mine material cost prediction problems, this paper combined the multiple regression model and RBF neural network, and established MRA-RBF coupling model to predict the material cost in coal mine. Coal cost forecast index system was established by selecting 8 variables from the aspects of natural factors, technical factors and management factors. The actual material cost in coal mine was predicted by using the established system. The maximum prediction error with MRA-RBF coupling model is 10.7951452%, the average prediction error is 5.45971%, and the minimum prediction error is only 0.3445817%. A better improvement is obtained compared with solo MRA and RBF. The results of study show that combined the linear fitting algorithm （MRA） and nonlinear fitting algorithm （RBF） to prediction of material cost for prediction of coal mine is more reliable, higher accuracy, and can be used in variable analysis. This model provides a new method for the prediction and control of coal mine material cost.

关 键 词：煤矿材料成本 多元回归分析 RBF神经网络 MRA-RBF耦合模 预测 成本控制 

分 类 号：F276.44[经济管理—企业管理]