Mining Actionable Knowledge Using Reordering Based Diversified Actionable Decision Trees

Actionable knowledge discovery plays a vital role in industrial problems such as Customer Relationship Management, insurance and banking. Actionable knowledge discovery techniques are not only useful in pointing out customers who are loyal and likely attritors, but it also suggests actions to transform customers from undesirable to desirable. Postprocessing is one of the actionable knowledge discovery techniques which are efficient and effective in strategic decision making and used to unearth hidden patterns and unknown correlations underlying the business data. In this paper, we present a novel technique named Reordering based Diversified Actionable Decision Trees (RDADT), which is an effective actionable knowledge discovery based classification algorithm. RDADT contrasts traditional classification algorithms by constructing committees of decision trees in a reordered fashion and discover actionable rules containing all the attributes. Experimental evaluation on UCI benchmark data shows that the proposed technique has higher classification accuracy than traditional decision tree algorithms.