The false positive and false negative rates are improved using user feedback on the misclassified information with incremental learning to dynamically adjust the rule weights.