Abstract:

Class prediction is an essential task in automatic speech recognition. The growing information amount and high dimensionality of data, relatively in number of characteristics (variables) and the number of samples (observations), make the application of many prediction techniques (e.g., logistic regression, discriminant analysis) non-significant and challenges. A proposed method to solve this problem is to employ dimension reduction statistical techniques. Successfully used in many areas and applied in statistical-related applications, common factor analysis (CFA) provides an efficient way for handling high-dimensional data. In this paper, the potential of CFA-based modeling in dimensionality reduction with multiclass target expression data is studied, and the power class prediction of linear discriminant analysis (LDA) in numeric data to develop a classification model is investigated. The proposed framework is effectively on the available experimental dataset, and are the results are interpreted. The efficacy of the suggested framework over the well known frameworks is also discussed.