Machine learning is one of the most active field today,and draws attentions and researches from various circles of society.It not only has been applied in some specific field,such as medical,text processing,image dealing,but also keeps its developing trends as a new star.Machine learning is based on data,and many problems around us may have a lot of factors that may be related,which makes the features become high dimensional.The high dimensional features are not only inefficient on training and predicting,but also have bad impact on learning algorithm for the unrelated and redundant features.Feature selection is an effective way to solve these problems.In this paper,I first introduce the process of feature preprocessing and feature selection,and then classify the feature selection from the two aspects of search strategy and evaluation criterion,and explain the method.We choose the support vector machine to evaluate the result of the feature selection,and give the introduction to the principle,the derivation,the concept of the kernel and the generalization method.Particle swarm optimization and cuckoo search can be applied as a random strategy to feature selection.Therefore,this paper will describe the principle and method of the two algorithms and lead them to the binary version according to the specific needs.An improved method is proposed for the binary cuckoo search algorithm after analyzing the search behavior of the binary cuckoo algorithm,which makes the algorithm have better searching direction and convergence precision.Based on this,the PSO-NCS algorithm is proposed.IN this method,the improved binary cuckoo search algorithm is combined with the PSO algorithm to make full use of the gathering performance of PSO when it’s converging and the global searching ability of CS.So that,the PSO-NCS algorithm not only can get out of the area of local best,but also can converging at a better result.Experiments are carried out on the PSO-NCS algorithm,the result seems to be better than some other algorithms.It is proved that,PSO-NCS algorithm has a better searching ability and quicker in convergence. |