Decission tree such as cart and chaid and their business application

Abstract

In term 3 of our PGP course, we got a chance to work on a real time analytics problem from one of the prominent Cement Manufacturing companies in India. One of the key drivers of the company’s business is to be able convert the contractors in the last mile, who are currently using cement from competitor brands to their brands. Considering a visit from the company sales executive to a contractor site for the purpose of conversion as a transaction, there is a huge repository of the data of such visits including in-depth data about the site and the result of the visit – Conversion or NonConversion of the contractor. The possible variables that could affect the conversion are the timing of visit to the contractor, the type of cement used currently by the contractor, the stage of construction, the site-requirement of cement etc., which are the constituents of each of the data points shared with us. The problem statement we had was to develop a model which will identify the contractor profiles to target based on variables stated above. The idea is to maximize the potential of the visits by focussing on profiles of contractors who would be more receptive to conversion. This formed the basis of our motivation to learn Classification Techniques which are Supervised Learning Algorithms that might be used to solve such problems, with binary class as outputs.