Feature Selection Based on Adaptive Genetic Algorithm and SVM
Min Chun Hu · Jisuanji gongcheng · 2009
Aiming at the question that the traditional method for discerning risk can not come true the real-time recognition of risk statue in the shield tunneling constructing process,this paper proposes a feature selection method which combines Adaptive Genetic Algorithm with Support Vector Machine(AGASVM).It is used to filter a pivotal feature subset which is super correlative with risk of constructing quality.Experimental result shows that the pivotal feature subset selected by AGASVM can make the classification accuracy higher when it is used in the real-time recognition of risk statue.The dimension of pivotal feature subset is obviously smaller than the one of original factors set,and the most of pivotal features are the same as the ideas of domain experts.