California Housing Price Prediction

Project Summary

This project implements a machine learning solution to predict housing prices in California using the California Housing dataset. The solution includes data preprocessing, exploratory data analysis, model training with hyperparameter tuning, and a user-friendly Streamlit web application for making predictions.

Features

• Data Analysis: Comprehensive EDA with visualizations
• Models Used: Linear Regression and Random Forest implementations
• Hyperparameter Tuning: GridSearchCV for optimal model performance
• Criteria for Feature Extraction: Analysis of critical factors
• Interactive Web App: Streamlit interface for user-friendly UI

Installation & Usage

1. Clone the repository

git clone https://github.com/Antrita/Predicting-House-Prices-using-Machine-Learning.git
cd california-housing-prediction

2. Installing Dependencies

pip install -r requirements.txt

3. Run the Jupyter Notebook

jupyter notebook housing_prediction.ipynb

4. Launch the Streamlit application

streamlit run app.py

Model Performance

Metrics Explanation

• MAE (Mean Absolute Error): Average prediction error in dollars
• MSE (Mean Squared Error): Average squared prediction error
• R² Score: Proportion of variance explained (closer to 1 is better)

Results

Model	MAE	MSE	R² Score
Linear Regression	$53,320	0.5559	0.5758
Random Forest	$32,754	0.2554	0.8051
Random Forest (Tuned)	$32,681	0.2540	0.8062

Hyperparameter Tuning Details

GridSearchCV Parameters

The Random Forest model was optimized using GridSearchCV with the following parameter grid:

param_grid = {
    'n_estimators': [50, 100, 200],      # Number of trees in the forest
    'max_depth': [10, 20, None],         # Maximum depth of trees
    'min_samples_split': [2, 5],         # Minimum samples required to split a node
    'min_samples_leaf': [1, 2]           # Minimum samples required at a leaf node
}

Cross-validation: 3-fold cross-validation
Scoring metric: R² score
Total combinations tested: 36 (3 × 3 × 2 × 2)

Best Parameters Found

After hyperparameter tuning, the optimal parameters were:

• n_estimators: 100-200 (varies by run)
• max_depth: 20 or None
• min_samples_split: 2-5
• min_samples_leaf: 1-2

Feature Importance Analysis

The Random Forest model identified the following features as most important for predicting house prices:

1. MedInc (Median Income): ~40% importance - The strongest predictor
2. AveOccup (Average Occupancy): ~15% importance
3. Longitude: ~15% importance - Location matters
4. Latitude: ~14% importance - Location matters
5. HouseAge: ~7% importance
6. AveRooms: ~5% importance
7. Population: ~3% importance
8. AveBedrms: ~1% importance

Data Preprocessing Details

Train-Test Split

• Training set: 80% of data (16,512 samples)
• Test set: 20% of data (4,128 samples)
• Random state: 42 (for reproducibility)

Feature Scaling

• StandardScaler applied to features for Linear Regression
• Random Forest used raw features (no scaling required)

Streamlit App Screenshot and example

Given the UI inputs shown:

• MedInc: 3.0 → $30,000 median income
• HouseAge: 10.0 → 10-year-old homes
• AveRooms: 6.0 → 6 rooms average
• AveBedrms: 2.0 → 2 bedrooms average
• Population: 1000 → Medium density
• AveOccup: 5.0 → 5 people per household (higher than typical)
• Latitude: 34.0 → Southern California
• Longitude: -118.0 → Los Angeles area

Model Output: $179,758.50 (±$50,000)

5. Instructions for Google Colab

To run this in Google Colab:

1. Create a new Colab notebook
2. Copy the Jupyter notebook code into cells
3. Run all cells sequentially
4. Download the generated model files (best_rf_model.pkl, scaler.pkl)
5. Create app.py locally with the Streamlit code
6. Run the Streamlit app locally with the downloaded model files

Bonus Features Implemented

✅ Feature importance visualization ✅ GridSearchCV hyperparameter tuning ✅ Model saving with joblib ✅ Comprehensive documentation ✅ Interactive Streamlit UI with metrics displayed