Housing Price Prediction Project

This project aims to predict housing prices using various regression models. The dataset used is from a housing dataset containing information about different houses' geographical location, median age, number of rooms, population, and other features.

Dataset

The dataset contains the following columns:

• longitude
• latitude
• housing_median_age
• total_rooms
• total_bedrooms
• population
• households
• median_income
• median_house_value
• ocean_proximity

Data Preprocessing

1. Handling Missing Values:

   • Dropped rows with missing values in the total_bedrooms column.

2. Log Transformation:

   • Applied log transformation to total_rooms, total_bedrooms, population, and households to reduce skewness.

3. One-Hot Encoding:

   • Converted categorical ocean_proximity column to dummy variables.

4. Feature Scaling:

   • Scaled the features using StandardScaler.

Exploratory Data Analysis (EDA)

• Plotted histograms for numerical columns before and after log transformation.
• Created a heatmap to visualize the correlation between features.
• Used scatter plots to understand geographical distribution.

Model Training and Evaluation

Models Used

1. Linear Regression
2. Ridge Regression
3. Lasso Regression
4. Random Forest Regressor
5. Gradient Boosting Regressor
6. XGBoost Regressor
7. Decision Tree Regressor

Model Evaluation

The models were evaluated using Root Mean Squared Error (RMSE) and R² score.

Model	RMSE	R² Score
Linear Regression	67775.13	0.664
Ridge Regression	67742.03	0.664
Lasso Regression	67699.92	0.665
Random Forest	48516.43	0.828
Gradient Boosting	56816.72	0.764
XGBoost	49541.45	0.821
Decision Tree	67766.99	0.664

Random Forest Regressor performed the best with the lowest RMSE and highest R² score.