Improved Adaboost Algorithm Apply On Gene Expression Data

Cancer diagnosis will be a major concern problem in the future.At present,cancer diagnosis is mainly focused on morphological,and the exact diagnosis is often in advanced stage of cancer.At this time,the cancer treatment is limited and the cure rate is low.The earlier diagnosed with cancer,the higher the cure rate.Early diagnosis is the key to cancer treatment.This paper summarizes the characteristics of cancer genes,diagnoses cancer from the genetic level,and hopes to achieve early diagnosis of cancer.The following aspects of the study are:(1)This paper presents an Adaboost algorithm based on AGA-ELM(Adaptive Genetic Algorithm for Extreme Learning Machine)for gene expression data classification,called Adaboost-ELM.The base classifier of this algorithm is Extreme Learning Machine(ELM).ELM has the advantages of high learning efficiency and generalization ability.But,the classification result of ELM is affected by random initialization of input weight matrix and the hidden layer bias.In order to solve those problems,this paper optimizes the input weight matrix and hidden layer bias of ELM by adaptive genetic algorithm.The problem of ELM random error is improved,and the stability of ELM algorithm is improved.(2)We proposed a hybrid Adaboost based on genetic algorithm for gene expression data classification,called Adaboost-GA.The base classifiers are decision groups composed by weak classifiers,including the K nearest neighbor(KNN)algorithm,Na?ve Bayes classifier and decision tree.The purpose of this paper is to summarize the advantages and disadvantages of different classifiers,and to improve the overall classification accuracy of gene expression data.After promoting the weak classifiers to strong classifiers,a genetic algorithm is employed to optimize the weight of each weak classifier.Adaboost-GA achieves the aim of improving classification accuracy and classification stability.