Abstract:

The nearest neighbor search algorithm is a simple method with high accuracy in text categorization, but it usually requires large amounts of calculation in the classifying process. To overcome this disadvantage, a twoclass text categorization method is proposed based on the nearest subspace search. It extracts a feature subspace from samples in the same class, and maps it to a point in a higher dimensional space, in which the classifying process is carried out by nearest neighbor search. Experiments on Reuters-21578 data sets show that the proposed method can effectively improve the performance of nearest neighbor search in text categorization, achieving a higher precision, recall rate, and F1 values.

Key words: text categorization;nearest subspace search;nearest neighbor search