How does PHPP support entropy discretization?

[Anders L Madsen]

The entropy discretization process aims at creating intervals such that each interval contains N/n values where n is the number of intervals and N is the number of cases. The aim is to create a uniform prior distribution on the continuous variable.

First the values are sorted in increasing order. Secondly the algorithm works from low to high values splitting an interval on the average between two neighbor values belonging to two different intervals. Each interval is created by counting values from low to high and  specifying the end point of the interval after N/n values. The end point of the current interval (which is the start point of the subsequent interval) is determined as the average of the last value in the current interval and the first value of the subsequent interval.