【作者】 孙祖妮；

【导师】 张秀媛；

【作者基本信息】 北京交通大学 ， 运输与物流， 2012， 硕士

【摘要】 铁路货物运输是铁路运输的重要组成部分,合理的控制货运成本可以有效地降低铁路运输总成本,而加强货运成本管理的重要前提就是准确及时地进行货运成本预测。选择合理的预测方法,可以提前为铁路运输企业提供有效的成本信息,为企业应对复杂多变的市场环境提供保障。成本预测方法很多,但并不是所有预测方法都符合铁路运输成本的特点,适合我国铁路运输的实际情况。本文通过比较不同预测方法得到的铁路货运成本预测结果,确定一个符合铁路运输成本特点的预测方法。通过分析我国铁路运输成本的构成及影响因素可以得出,在铁路运输成本支出中间接成本占较大比重,而且运输成本数据具有较强的非线性,在对运输成本进行预测之前,既要选择恰当的间接费用分配指标合理地分配客货运成本,又要充分考虑铁路运输成本预测自身特性。基于以上两点考虑,选择作业成本法和BP神经网络相结合的方法对铁路货运成本进行预测。首先,从作业角度出发,利用成本动因理论,选取对货运成本影响较大的成本动因作为分配间接成本的指标,并结合定性分析和定量分析利用聚类分析法合并成本动因,再利用BP神经网络很强的学习能力、容错能力以及非线性映射能力,解决运输成本样本采集不精确和成本数据呈非线性关系的问题。以所选成本动因为输入变量,在MATLAB软件中反复进行信号的正向传播和误差的反向传播过程,不断学习训练,存储学习结果,获得预测结果。并将结果与灰色系统预测方法所得的预测结果进行比较。最后以某铁路运输企业货运成本预测为例,分别用成本动因BP神经网络预测方法和灰色系统预测方法进行预测,并分析。结果表明,通过作业成本法选择成本动因,然后利用BP神经网络模型进行成本预测的方法比灰色系统预测法更适合用于国铁路货运成本预测。更多还原

【Abstract】 Railway freight transportation is one of the most important parts of railway transportation. To extend effective control for cost saving can reduce the total railway transportation cost. On the premise of that is to forecast cost accurately and timely. So choosing a reasonably prediction method can not only provide effective cost information, but also make the company do well in the complicated and volatile economic environment. There are a lot of prediction methods, but not all of them fit for railway transportation cost and the situation of our country’s railway transportation. In this paper, we compare the forecast results of some prediction methods to find a method which is more suit our railway transportation.By analyzing the parts of the railway transportation cost and the influencing factors, we can get that the indirect cost has a large proportion in the transportation cost, and between the sample data of the transportation cost has a non-linear relationship. Before estimating the cost we should choose some reasonable distribution targets for the freight cost and consider the non-linear relationship of the cost. Based on the two reasons, we forecast the freight cost with ABC and BP neural network. ABC can consider all the activities in the transportation, and distribute the activities to the cost reasonably. In addition, BP neural network has Strong learning ability and fault-tolerant ability, it can solve the problems which the other methods cannot. Then using MATLAB software trains the network repeatedly to get the forecast results we need.At the last of the paper, the paper estimates a railway company’s freight cost with two methods. They are cost driver-BP neural network and grey system. Comparison of the predicted results, we get that the cost driver-BP neural network have an ideal result, it can be an ideal method for the freight cost estimation.更多还原