This Drug Classification project utilizes a Decision Tree Classifier to accurately predict prescribed medications based on patient health profiles. The implementation features a modular scikit-learn pipeline with a ColumnTransformer to handle various data types.
Dataset Used: https://www.kaggle.com/datasets/prathamtripathi/drug-classification
My Solution
I chose a Pipeline approach to keep the code clean and prevent data leakage.
The logic was to split the features based on their type.
Since BP and Cholesterol have a natural order, I used ordinal encoding.
For numeric values like the sodium-to-potassium ratio, I used scaling to ensure they are on the same level.
I limited the tree size to 6 leaf nodes to make sure the model stays simple and easy to interpret.
Results and Discussion
The model achieved 100% accuracy on the test set.
Performance Summary
The confusion matrix shows perfect precision and recall for all five drug classes (A, B, C, X, and Y).
