Statistical feature extraction for machine learning-based text mining

Abstract

Large scale databases are more common in many application areas of data mining. Most of them contains an incredible amount of information in text format, thus as the volume of electronic information grows, so does its complexity to analyze it and understand it. In this thesis, various types of statistical feature extraction and classification methods are introduced, and the performances of text classification for the benchmark data set Reuters-21578 are compared. It is also suggested the possible improvements of text mining methods through the analysis of simulation results.