A Novel Tree Based Classification

Classification is a data mining (DM) technique used to predict or forecast the unknown information using the historical data. There are many classification techniques. ID3 is a very popular tree based classification algorithm for a categorical data which does not support continuous data. Attribute selection process plays major role in building a classification tree model. Attribute Selection in ID3 is based on the information gain. For a continuous attribute there is an extended version of ID3, called C4.5, which uses information gain ratio for discretization. Class-attribute interdependency Relation (CAIR) and Class-attribute interdependency maximization (CAIM) are different discretization schemes. CAIR gives better relationship between attribute and class then information gain and CAIM is a better discretization scheme then the one used in C4.5. In this paper A Novel Tree Based Classification Algorithm is proposed. It generates classification tree model in the similar manner as ID3. It uses CAIM based online discretization and attributes selection process. The work is extended with CAIR based attribute selection process. The results are tested on different datasets, which have shown the improved results.