Post Mining of Diversified Multiple Decision Trees for Actionable Knowledge Discovery

Abstract

Most data mining algorithms and tools when applied to industrial problems such as Customer Relationship Management, insurance and banking they stop search at producing actual applicable knowledge. Unlike these models, actionable knowledge discovery techniques are useful in pointing out customers who are likely attritors and loyal. However, actionable knowledge discovery techniques require human experts to postprocess the discovered knowledge manually. Postprocessing is one of the actionable knowledge discovery techniques which are effective in decision making and overcomes considerable inefficiency which leads to human errors that are inherent in the traditional data mining systems. Hence, decision trees are postprocessed which suggest cost effective actions in order to maximize the profit based objective function. In the proposed approach, an effective actionable knowledge discovery based classification algorithm namely Actionable Multiple Decision Trees (AMDT) is developed to improve the robustness and classification accuracy and tests are conducted on UCI German benchmark data.