Research Of Feature Selection Algorithm Based On Cuckoo Search

With the rapid development of science and technology, a mass of data and informationgenerated in every field in our daily life. These data and information really provide us withmuch convenience, they also bring some trouble to us. Sometimes we can hardly get what wewant from them. At the same time, more and more data is in high demension and low samplesize with the expansion size in data. The heavy noise in the feature space makes it harder forus to study and acknowledge the essence of the problems.Feature selection came into being in such background. We can eliminate many uselessfeatures by selecting an efficient subset from the original feature space. Thus, it’s easier for usto solve the problem and feature selection is widely used in real life. Many effectivealgorithms appeared with the development of feature selection. Cuckoo search algorithm(CS)is one of the new methods. Cuckoo search algorithm came from the simulation of thebehaviors of cuckoos when they lay eggs. The algorithm has the characteristics of lessparameters, fast convergence and so on.This paper brings in the thought of the quantum algorithm on the basis of the originalcuckoo search algorithm and binary cuckoo search algothm. A new feature selection methodbased on cuckoo search algorithm is presented.The main contents of this paper are as follows:(1)This paper introduces the background and current state of feature selection and alsogives the basic framework of feature selection. Several feature selection methods arepresented in different standard of classification.(2)This paper represents the origin of cuckoo search and gives a detailed description ofcuckoo search algorithm and binary cuckoo search algorithm.(3)In order to increase the efficiency of cuckooc search algorithm, this paper presents theidea of quantum algorithm and provides the derivation. A binary cuckoo algorithm based onquantum algorithm is proposed and the original cuckoo algorithm is improved. The improvedalgorithm no longer iterates towards random direction. It will move towards the optomalsolution with the step size, which comes from Levy flight. Thus, the improved algorithmconverges faster and reduces the total times of iteration. (4)The improved algorithm is used to solved the knapsack problem and feature selectionin the microarray dataset. SVM is used to evaluate feature subset which is selected from theoriginal dataset. The paper gives an introduction of microarray technology and microarraydataset along with the fundamentals of SVM. The paper gives a comparation with otherstate-of-art methods and proves that the new algorithm has an good performance.(5)In the end of the paper, a summary of the paper is presented and a few thoughts whichcould be considered in the future are also presented.