Chinese Text Dimensionality Reduction Based On Factor Analysis

With the fast development of networks and information technology, resulting in a largenumber of text information that can be mining, The common classification algorithms bringhigh computational complexity and also cannot produce good classification effect of corpusnoise. The text effective dimensionality reduction can not only improve the computing speed,and also can improve the efficiency and effectiveness of the classifier, therefore theimplementation of the text dimensionality reduction is very necessary..The thesis majors research and implementation of the factor-analysis algorithm, andprovides improvements of it. Factor-analysis as an effective means of dimensionalityreduction has been taken attention to people which bring many new methods for it. Manymethods are based on the covariance method for conversion. The use of the matrix orthogonalto rotate, resulting in the compression of information.The thesis also do the researches in the common dimensionality reduction of the Chinesetext.It discuss the classification of the key technologies, including the basis for building avector space, calculating weights, classification methods, etc. And it also analyzes thecharacteristics of the Chinese text. It describes the implementation process of the text-basedclassification algorithm dimensionality reduction. The last experiment with the classificationand comparison of factor-analysis evaluated the effectiveness and feasibility of the algorithm,and analysis of the experimental results and their reason has been taken.Classification and comparison of experimental results show that: factor-analysis methodsbased on KNN classification enhancement effects more obviously. Factor-analysis improvesSVM-based Classification. It can enhance the results of classification for sparse matrix ofnoise effect.