| 摘要: |
| 针对成本在经济学系统中变化的非对称，与其影响因素之间的非线性关系，提出采用神经网络分位数回归法来研究成本与各影响因素之间的联系，并进行成本预测；该方法不仅可以通过神经网络结构模拟经济系统中非线性关系，还可以通过分位数回归功能揭示各影响因素对成本整个条件分布的影响规律；通过实证结果分析，神经网络分位数回归模型相较于OLS回归模型和分位数回归模型其预测精度更高，且揭示了各因素的影响规律，所以神经网络分位数回归模型的分析结果更科学，更有价值，更有助于相关管理决策者进行成本分析、控制和管理。 |
| 关键词: 分位数回归 神经网络 成本预测 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Research on the Cost Prediction of the Industry Based on Quantile Regression Neural Network |
|
MENG Xiao lu
|
| Abstract: |
| Cost is an important factor affecting the survival of every enterprise. So， how to manage the cost and take advantage of the cost information is very important. For the change of cost being asymmetrical in the economic system and having nonlinear relation with its influence factors， the quantile regression neural network （QRNN） method is used to study the relation between cost and influence factors and predict cost. The QRNN method can not only simulate the nonlinear relation in the economic system through the neural network structure， but also can reveal the influence rules of each factor on the whole conditional distribution of the cost by quantile regression approach. By empirical analysis， the QRNN model has higher prediction accuracy by comparing the OLS regression model and the quantile regression model， and reveals the influence rules of various factors. So， the analysis results of QRNN model are more scientific， more valuable and more helpful to the related decision makers for cost analysis， control and management. |
| Key words: quantile regression neural network cost prediction |
