Automated learning of mean mixture normal exponential factor analysis models

Factor analysis (FA), as an unsupervised data miningtechnique, allows us to reduce technique in a variety ofscientific fields, especially in psychology and educationaltesting. In seeking the most appropriate number of factors qof a FA model, a two-stage procedure is commonlyimplemented by firstly carrying out parameter estimationover a set of prespecified numbers of factors, and thenselecting the best q according to certain penalizedlikelihood criteria. When the dimensionality of data growshigher, such a procedure can be computationallyprohibitive. To overcome this obstacle, we develop anautomated learning scheme, called the automated meanmixture normal exponential FA (AMMNEFA) algorithm, toeffectively merge parameter estimation and selection ofinto a one-stage algorithm.

کلیدواژه ها:

ECM-algorithm ، mean mixture normal distribution ، skewness ، Factor analysis ، heavy-tail distribution