Research On Cancer Gene Microarray Classification Method

The cancer is complex hereditary disease, have huge harmfulness and multiple. Ithad become a significant disease for influencing the human health, its pathogenesis isclosely linked to genes fundamentally. Genes microarray classification techniques canhelp human to find the essential diversity of the genes between normal cells anddisease cells, understand the pathogenesis of tumour and distinguish the carcinogenicgenes, It has very far-reaching for the cancer’s clinical diagnosis and cure. Asmicroarray data is characterized by high-dimensional noisy samples and traditionalstatistical methods is difficult to achieve its classification, In order to solve the aboveproblems, This article analysis the microarray classification techniques in-depthly, and based on it to carry out the study. The main contents are as follows:Decision Tree and Large Coverage Rule microarray classification techniquesembody relativity of genes, but still lack stability and algorithms convergence areslow to result in producing numerous redundancy classification rule. This paper isbased on genetic programming (GP), it proposes an approach called “Best RuleGenetic Algorithm (BRGA)” for optimizing classification rule, gain bestclassification rule set. This algorithm can adjust relevant parameter of classifiermodel, substantially improve performance of classification on the basis of increasingappropriate iteration, therefore it has considerable flexibility and intelligibility. Theperformance of the proposed approach is evaluated using six gene expression data setby simulation. From the result, it is found that the proposed approach reducesComputational Complexity and redundancy on the basis of good classificationaccuracy and stability than the other approaches reported in the other literature.Traditional Emerging Patterns microarray data classification method employeddiscrete information entropy process or other complicated algorithms wiping offinsignificance noise genes for obtaining the EP patterns at the selecting genes part andobtained the best discriminatory gene attributes. and then produced EP classificationmodel to predict unknown sample. But these methods have more complicatedcalculating and more pay expenses and not being apt to be comprehended. The paperpresent a Function Jumping Emerging Patterns (F_JEP) algorithm. It definemicroarray data as Function form in order to express easily. Being based on discretemethod for equivalent width way it employ a very simple aequilatus breakpoint search way to partition microarray gene attribute value and obtain significant classificationgenes. and then it produce F_JEP Patterns.The performance of the proposed approach is evaluated using six geneexpression data set by simulation. And F_JEP compared with three famous algorithms(NB,IB,C4.5) at the classification performance. From the result, it is found that theF_JEP algorithm precede the three algorithms on the cancer microarray datasetobviously. At the same time BRGA algorithm and F_JEP algorithm reduceComputational Complexity and operation pay expenses, remove redundancy on thebasis of good classification accuracy and stability than the other approaches reportedin the other literature.Both BRGA algorithm and F_JEP algorithm are efficient and flexible, they aremicroarray data classification algorithms with the strong expansibility. Because ofboundedness of the experiment condition and the biology development. the algorithmsstill need being improved on and complete. we hope that they are applied widely inthe fields for the biology and clinical medicine.