A Novel Scheme for Term Weighting in Text Categorization: Positive Impact Factor

The exploitation of Data Mining and Knowledge discovery has penetrated to a variety of Machine Learning Systems. A very important area in the field of Machine learning is Text Categorization. Feature selection and Term weighting are two important steps that decide the result of any Text Categorization problem. In this paper we focus our research on effective term weighting and propose a novel Term weighting approach i.e. Positive Impact Factor (PIF). PIF is a supervised variation of traditional term weighting models. The idea behind PIF scheme revolves around the assumption "Positive impact of a feature to a category can be used to calculate its negative impact for other categories." To examine our weighting scheme we used the dataset Classic 3 from Cornell, which has documents in 3 predefined categories. Results of our experiment and comparison with existing methods such as Binary, TF, TF-IDF, TF-RF, TF-CHI2 etc show remarkable improvement in accuracy with a significant reduction of computational cost.