California Housing Price Analysis

Overview

This project analyzes the California Housing dataset to understand the key factors that influence housing prices and to build predictive models.

Objectives

• Explore relationships between features and housing prices
• Perform data cleaning and preprocessing
• Build and compare machine learning models

Data

The dataset contains information from the 1990 California census, including:

• Median income
• Housing characteristics (rooms, population, households)
• Location (ocean proximity)
• Median house value

Methods

• Data cleaning (handling missing values)
• Exploratory Data Analysis (visualizations)
• Feature engineering (rooms per household)
• One-hot encoding for categorical variables
• Model building:
  • Linear Regression
  • Random Forest Regressor

Results

• Linear Regression R²: ~0.65
• Random Forest R² (test): ~0.83

The Random Forest model performed significantly better, indicating that housing prices depend on non-linear relationships between features.

Key Insights

• Median income is the strongest predictor of housing prices
• Houses near the ocean tend to be more expensive
• The dataset contains a price cap at $500,000, which may affect analysis

Tools Used

• Python
• pandas
• matplotlib
• scikit-learn

Conclusion

This project demonstrates how data preprocessing, feature engineering, and model selection impact predictive performance. More complex models like Random Forest can better capture real-world patterns compared to simple linear models.