Application Of Evolutionary Optimization Based Gaussian Mixture Models To Intelligent Data Analysis

Gaussian mixture model (Gaussian Mixture Models, GMM) is an important method of machine learning, which aims to establish a probability model for the target data set. In addition, GMM optimization strategy is the key to improving the accuracy of the model; The process includes an accurate estimate of the number of GMM components and the corresponding parameters.Given the GMM standard parameter estimation strategy——the Expectation-Maximization(EM) algorithm, often require more accurate initialization and easily gets trapped in local maxima, we introduce based on Akaike information criterion (AIC) and evolution algorithms instead of EM algorithm to optimization GMM, the evolution algorithm is based on swarm intelligence theory, through inter-group cooperation and competition between individuals to produce new individuals evolve new populations, evolution algorithm, the initial solution is less demanding, but simple and efficient global optimization has become widely used algorithms. In the optimization process, but because of the particularity of GMM covariance matrix, their uses for GMM estimation have been limited to constrained models using identity or diagonal covariance matrices. In this paper, we used a novel parameterization for arbitrary covariance matrices that allow independent updating of individual parameters while retaining validity of the resultant matrices. The accuracy of the modeling on synthetic and real data sets show that and evolution algorithm and AIC based GMM for the modeling and analysis of intelligent data has a good effect.