Calculating certainty in decision trees: Certainty vs Entropy.

Code

• github.com/alexwareb9/cid3
  none★ 6

Abstract

This research introduces a certainty function, to replace the entropy function used to calculate information gain when building decision trees, in algorithms like ID3[1], C4.5[2] and C5.0. The new function is simpler than the entropy function and calculates certainty more accurately, potentially consuming less time in the process. I will also introduce another formula that help deal with causal analysis. It is the certainty-raising inequality. This formula can be derived from the original certainty function, and provide insights into the cause-effect relations in the data.

Reproductions