Abstract: This research paper is about constructive induction of data mining (DM) which is used to cope with inadequacy of attributes. Constructive induction is a process of learning concept description that represents best hypothesis in the space. Inductive learning algorithms are increasingly being pressed into service, as data mining and knowledge discovery tools, to detect patterns or regularities in large amounts of data. A major limitation of conventional learning algorithm is that the descriptions they build such as decision trees, decision rules and bayesan nets employ only terms selected from among those explicitly provided in the data.

Keywords: Over precision, Attribute Interaction, Irrelevant attribute, Data driven constructive induction systems, SVD (Singular Value Decomposition), Knowledge-Driven constructive induction systems.