Problem 4. You have been hired as a data mining consultant for an insurance company. Your first task is

to apply clustering to segment the customers who are insured with the company. The customer data set

contains only 10 categorical attributes (gender, marital status, occupation, highest education level, etc).

1) You have seen how entropy is used in decision tree classifiers to split data points with categorical

attributes into homogeneous subgroups, so you decided to develop an entropy-based divisive

hierarchical clustering algorithm that can handle categorical attributes. The algorithm starts with 1

cluster containing all the points and chooses an attribute that gives the highest information gain to

partition the instances into its smaller clusters. It repeats this partitioning step on each cluster until

every cluster contains only a single point. You present the idea to your supervisor, who immediately

shoots it down. Explain why such an approach conceptually will not work.