| Feature | ID3 | C4.5 | CART |
|---|---|---|---|
| Developer | Ross Quinlan | Ross Quinlan | Breiman et al. |
| Splitting Criterion | Information Gain | Gain Ratio | Gini Impurity (Classification), Mean Squared Error (Regression) |
| Data Types Supported | Categorical | Categorical & Continuous | Categorical & Continuous |
| Pruning | No | Post-pruning | Cost-complexity pruning |
| Output | Multi-way splits | Multi-way splits | Binary splits |
| Regression Support | No | No | Yes |
| Handles Missing Data | No | Yes | Yes |
| Computational Efficiency | Fast | Moderate | Moderate |
| Bias Towards Features | Yes, towards features with many categories |
No, uses Gain Ratio to normalize | No |
| Strengths | Simple and easy to implement | Handles continuous data, avoids bias, post-pruning |
Supports both classification and regression; robust |
| Weaknesses | Overfitting prone, categorical data only |
Slower than ID3, no regression |
Only binary splits, more computationally intensive |
| Applications | Simple classification tasks (e.g., weather prediction) |
Complex classification tasks (e.g., disease diagnosis) |
Both classification and regression (e.g., fraud detection, house price prediction) |