In this lesson, we will discuss about different parameters present to control a Decision Tree (DT). Two features, max_depth and min_samples_leaf have a greater importance than other parameters. We will further see how we first tune max_depth parameter and then move to tuning other parameters will help. Finally, a dataframe is created with all possible combinations of max_depth, min_sample_leaf and the auc score corresponding to them. These results are visualized using a heatmap by pivoting the dataframe to easily determine the best possible max_depth and min_samples_leaf combination. Finally, the DT is retrained using the identified parameter combination. DT so trained is viewed as a tree diagram.
Add notes from the video (PRs are welcome)
- iterating to find optimal parameter settings
- creating the heatmap with seaborn
|
The notes are written by the community. If you see an error here, please create a PR with a fix. |