A Novel Text Classification Approach Based on Enhanced Association Rule

The current research on association rule based text classification neglected several key problems. First, weights of elements in profile vectors may have much impact on generating classification rules. Second, traditional association rule lacks semantics. Increasing semantic of association rule may help to improve the classification accuracy. Focusing on the above problems, we propose a new classification approach. This approach include: (1) Mining frequent item-sets on item-weighted transactions; (2) Generating enhanced association rule that has richer semantics than traditional association rule. Experiments show that new approach outperforms CMAR, S-EM and NB algorithms on classification accuracy.