Feature subset selection in SOM based text categorization

Feature selection, as a preprocessing step to machine learning, is effective in reducing dimensionality, removing irrelevant data, increasing learning accuracy, and improving result comprehensibility. In this paper we describes several methods for feature subset selection on large text data. The experimental comparison of the described methods will be given on the collected data from the Web, focusing on aggressive dimensionality reduction. In our experiments the well known self-organizing map (SOM) neural net-work was used as text categorization algorithm. We will also, investigates the effect of dimensionality reduction by this methods on the performance of text categorization using both entropy and f-measure as evaluation measures.