Regression Learning Curves

Learning curve comparison of KNN, Decision Tree, and Linear Regression regressors on an OpenML dataset.

Overview

This project compares regression models using learning curves computed on OpenML dataset 541. Categorical features are one-hot encoded before training. Task 1 plots learning curves for KNN regressors with k = 3, 5, and 7. Task 2 tunes KNN and Decision Tree regressors via GridSearchCV, then plots learning curves for the tuned KNN, tuned Decision Tree, and a baseline Linear Regression model. RMSE values at the largest training size are printed in a table for each task.

Requirements

• Python 3.8+
• scikit-learn >= 1.0
• matplotlib >= 3.5
• numpy >= 1.21

Installation

pip install -r requirements.txt

Project Structure

regression-learning-curves/
├── compare_regressors.py   # Learning curve comparison script
├── requirements.txt        # Python dependencies
├── .gitignore
└── README.md

Usage

python compare_regressors.py

Steps performed:

1. Downloads OpenML dataset 541 and one-hot encodes categorical columns.
2. Task 1: plots learning curves for KNN with k = 3, 5, 7; saves task_1_knn_with_different_k_values.png.
3. Task 2: runs GridSearchCV for KNN and Decision Tree, then plots learning curves for the tuned models and Linear Regression; saves task_2_tuned_models_and_linear_regression.png.
4. Prints an RMSE summary table for each task.

Results

Two PNG plots and two RMSE tables are produced:

File	Description
task_1_knn_with_different_k_values.png	KNN learning curves for k = 3, 5, 7
task_2_tuned_models_and_linear_regression.png	Tuned KNN, tuned DT, and LR learning curves

RMSE values at the maximum training set size are printed to stdout beneath each plot.

Author

Biswajeet Sahoo

License

MIT License