基于极大熵聚类的工程项目风险预警模型

该方法针对当前大部分工程风险预警模型只能报警,不能预测的现状,提出了基于熵最优化的工程项目风险预警方法.利用判别熵最小化选取项目风险预警指标值,通过特征选取找出那些最有效的特征,研究出一种新的聚类算法——极大熵聚类算法,极大熵聚类算法是以概率为比例将任一指标向量分配给所有码向量,而不是仅仅只分配给与之最近的码向量,该算法是C-均值算法的一种推广.最后用实例验证该模型,用此算法对预测结果进行分类,判断项目的风险状态.结果表明这种方法估计工程项目风险快捷有效,与实际情况基本一致,可以应用于工程分析. 

Early-warning of engineering-project risk model research based on maximum entropy clustering

Due to most of current early-warning of engineering risk only give a alarm,but cannot forecast,engineering-project risk early-warning based on entropy optimal model was proposed.Firstly,minimum J-divergence entropy was applied to extract the risk early-warning index and to find out the most effective feature by feature extraction.Then the calculating result was classified to judge state of project with a new clustering algorithm-maximum entropy clustering algorithm.Maximum entropy clustering algorithm allocated index vectors to all of the code vectors rather than the nearest code vector with a ratio of possibility.The algorithm is an improved c-means algorithm.Finally,the case was verified to,the results of forecast were classified by the algorithm to estimate the project's venture.The experiment results show that the improved algorithm can use to predict the project risk quickly and effectively at engineering analyses.The analytical results are basically identical with the actual situation.