Research Of Classification Based On Learning Automata

AdaBoost is a classical classification algorithm in the area of data mining,but prone to be easily overfitting in the environment with noise.In recent years,many researchers have put forward some improved algorithms.However,it's still a lot of work to be done with AdaBoost in the environment with noise.Considering of the learning automata could search the optimal value in the environment with noise,this paper do the research of AdaBoost based on learning automata.The first part is the analysis of AdaBoost,then in the next part an improved algorithm based on continuous action learning automata is proposed.In addition,considering of the convergence rate of continuous action learning automata,this paper gives an improved method based on stochastic point location.In this paper,the discription of the two algorithms is given,and the simulation experiment is carried out.Sastry has presented a kind of classifier based on learning automata which performs well on the environment with noise.But this algorithm could not deal with nonlinear data sets.To solve the problem,this paper gives the thought of classification on the nonlinear data sets based on ensemble learning.To balance the efficiency and accuracy,this paper mainly uses Boosting to improve the performace of original algorithm.Experimental results show that the method can achieve the aim to classify nonlinear data sets,and has a good performance in the environment with noise.