A density based aapproach to classification

This paper presents a novel method for classification, which is density based and makes use of the models built by the lattice machine (LM) [5, 7]. Density is a natural concept to use in clustering and the LM is a relatively new method for supervised learning developed in recent years. The LM approximates data resulting in, as a model of data, a set of hyper tuples that are equilabelled, supported and maximal. The method presented in this paper uses the LM model of data to classify new data with a view to maximising the density of the model. In order for the method to have wide applicability a measure of density is introduced for hyper tuples and relations.Experiments were carried out with both public and proprietary data. Experimental results show that our method outperforms the classification method in the LM literature and it is comparable to the C5.0 classification algorithm. It is also shown that our method works quite well in an application-stock market data mining.