Methods Of Lazy Bayesian Classification Based On Emerging Patterns

Classification is an important topic in data mining community. With the growth of data in volume and dimensionality, it is getting more and more challenging to build effective classifiers for large-scale databases. Methods for classification based on Emerging Patterns (EP) were proposed in order to classify large datasets efficiently. EP is a new kind of knowledge pattern, which can discover the inherent distinctions between different classes of data.In this paper, we firstly introduce the concept and basic technology about classification. And then detailedly present the basic concept about EP, efficient EP mining algorithms and effective methods to use EP for classification. Furthermore we throughly analyze the algorithms which EP-based classification that has been proposed. Finally based on these algorithms, we realize the classification algorithm of DeEPs and apply an improvement on DeEPs. We propose a new classification algorithm called Lazy Bayesian Classification based on essential Emerging Patterns (LBCeEP). The algorithm applies lazy learning strategy to reduce training set and adopts a special kind of EP for classification. Differing from the existing EP-based classifiers, LBCeEP uses Bayesian approach to measure the contribution of EP for classification. Moreover, LBCeEP can be self-adaptive to parameter. Our experimental results have shown that LBCeEP performs comparably with other state-of-the-art classification methods such as NB, TAN, C4.5 and DeEPs.