This project uses machine learning to predict house sale prices using the Iowa Home Prices Dataset. It was created as part of the Kaggle Intro to Machine Learning course.
📂 Project Structure
train.csv – Training data used to build the model
test.csv – Test data used to generate predictions
notebook.ipynb – Jupyter notebook containing all steps (data prep, modeling, evaluation)
submission.csv – Final predictions ready for Kaggle submission
🧠 Approach
Load & Explore Data – Selected key features like:
LotArea, YearBuilt, 1stFlrSF, 2ndFlrSF, FullBath, BedroomAbvGr, TotRmsAbvGrd
Split Dataset – Used train_test_split to create training/validation sets.
Train Model – Started with a DecisionTreeRegressor, then improved accuracy using RandomForestRegressor.
Evaluate Model – Used Mean Absolute Error (MAE) to compare models.
Train on Full Data – Trained the best model (RandomForestRegressor) on all available data.
Make Predictions – Predicted on the test.csv dataset and saved results as submission.csv.
📊 Results Model Validation MAE Decision Tree (Default) ~28,000 Decision Tree (Tuned) ~24,000 Random Forest ~18,000 ✅ 🚀 How to Reproduce
Clone this repo or open the notebook in Kaggle.
Run all cells in order to generate submission.csv.
Go to Data tab → submission.csv → Submit to send results to the competition leaderboard.
🏆 Competition Link
🔗 Kaggle Competition – Home Data for ML Course