基于数据挖掘的施工危险识别注意投入与分配研究*

为分析施工情境中危险识别注意资源动态投入分配规律，基于动态时间规整算法，挖掘危险识别注视轨迹序列，以表征注意资源投入分配变化，并采用k-means聚类、注视熵、Needleman-Wunsch全局序列对齐算法和统计等方法，深入挖掘注意资源在危险目标中投入和分配等时空变化规律。研究结果表明:当事人危险识别各阶段注意资源呈现从显著目标到高危目标的投入变化趋势，危险识别注意资源分配随情境复杂因素呈现零散、均匀的空间特征，分配无序程度提高。

Attention input and allocation of construction hazard identification based on data mining

In order to analyze the dynamic input allocation law of hazard identification attention resources in the construction situation,the gaze track sequence of hazard identification was mined based on the dynamic time warping algorithm to characterize the change in attention resource input allocation.The k-means clustering,gaze entropy,Needleman-Wunsch global sequence alignment algorithm and statistics methods were applied to deeply mine the spatiotemporal change in the input and allocation of attention resources on dangerous targets.The results showed that the attention resources in each stage of the client’s hazard identification presented a changing trend of input from significant targets to high-risk targets.The allocation of hazard identification attention resources presented the scattered and uniform spatial characteristics with the complex factors of situation,and the degree of disorder in allocation increased.