Demographic Customer Segmentation of banking users Based on k-prototype methodology

Abstract

With the ever-increasing population size, comes an ever-increasing diversity in tastes and preferences. Catering to each of these nearly 7 billion preferences individually is an unimaginable task. Whereas providing the same service to whole population would nullify the meaning of ‘preferences. This is where customer segmentation acts as a middle ground. Customer segmentation is a way to cater to tastes and preferences of groups of individuals rather than individuals itself. Although, the individuals in these groups might not have the exact same preferences, but they lie in the same ballpark, making them more similar to each other than the individuals of other groups. Segmentation is the first step in ‘targeted marketing’, which is followed my targeting and eventually by positioning. One way of performing said segmentation is by manually segregating customers one by one, be it by using MS Excel or any query language. But this way is very cumbersome and error prone, it is also very time inefficient. Therefore, machine learning algorithms are used for big data sets. This not only eliminates the above problems, but it also increases the scope of analysis through data manipulation and visualization. The most common machine learning algorithms used for customer segmentation are the unsupervised clustering algorithms out of which k-means is the most popular one. We are going to study a variation of this k-prototype and look at how it performs when it comes to customer segmentation.