An Improved Text Feature Selection Method for Transfer Learning

Text classification is an important technology in text mining, and widely used in the real world. Two-class classification problem plays an important role in text classification. Many practical problems, including web page classification and spam filtering, are essentially two-class text classification problems. In this paper, we use two-class text classification problem as a baseline, and improve the traditional method for text feature selection so that the method can be more appropriate for transfer learning. We apply the improved method for text feature selection with a few new data and much old data. After the test, the updated method is proven to have effectively improved upon the classification precision and rate of recall.