Abstract: Naive Bayesian classifiers have difficult problems involving getting labeled training datasets, and cost a lot of time to learn all samples again when new sample adds. Motivated by this fact, the paper presents an incremental learning method, and proposes a weighted naive Bayesian classification algorithm. All of them improve the performance of naive Bayesian classifiers at the expense of attribute weights, the attribute weighted parameters are directly induced from training dataset. Experimentally testing the algorithm using the UCI datasets recommended by Weka, the results show that the algorithm is feasible and effective.

Key words: naive Bayesian classifiers, attribute weights, incremental learning, training dataset