基于数据挖掘的地下燃气管道风险指标体系设计

针对风险模型的建立进行建模指标集的确定。鉴于指标数量多,相关关系错综复杂,先后采用了多种数据挖掘方法：专家调研法、最小均方差法、极小极大离差法逐步对指标进行筛选。并根据各指标对地下燃气管道影响的方式不同,指标间相关关系的差异确定了三层结构的风险模型指标体系,采用从三级、二级到一级指标的逐级量化方案。运用专家评分法在综合指标信息方面的优势,避开其由于指标集过大而导致评分的偏差,在小指标集内进行指标量化。将层次分析法运用到处理专家意见,确定指标权重的过程中。

Design for underground gas pipe-line risk factors system based on data mining

Firstly aimed at the establishment of risk models,the index set for model was confirmed.In view of large quantity of indexes and complex relation between them,many data mining methods were adopted early or late,which include Delphi,least averages square-difference and maximum or least square-difference.Taking into account the different ways of the every index effect on the underground gas pipelines and the diversity of relation among indexes,the risks model index system was adopted with three layers framework.The stepwise quantification project was adopted from third class,second class to first class.By application of specialist predominance in integration of index information,the situation that large quantity of index set leads to evaluation windage and quantify the indexes in small quantity of indexes can be avoided.An analytic hierarchy process in dealing with experts' opinions to confirm the index weights is also applied.