Abstract: Man-hour is affected mixed type factors,and man-hour function has the property of non-convex and non-continuity in process industry.To solve the low prediction precision of standard operation time,a man-hour estimation method based on decision tree and model tree was proposed,which was used to fit the non-linear relationship.For the data set of mixed attribute types,the nominal attributes were used to build the part of tree based on decision tree method,and the model tree method was used to construct each sub-tree.A case was used to demonstrate that the proposed method could reduce the estimation deviation by 19.4% on the basis of current method,and reduce the discrepancy between plan and execution.

Key words: man-hour estimation, process industry, decision tree, model tree

摘要： 为解决标准生产工时预测精度低的问题,提出基于决策树和模型树的作业工时预估方法,用于拟合非线性关系。对于可获取的由混合属性类型构成的数据集,先基于决策树法用标称属性完成部分树的构建,然后在各分枝的数据集上采用模型树法构建子树,在叶节点处给出线性模型。通过实例证明,所提方法在当前基础上能使预估偏差降低19.4%,显著缩小计划与执行之间的偏差。

关键词: 工时估计, 流程工业, 决策树, 模型树