Real Estate House Prices Prediction

This repository contains a Jupyter Notebook that demonstrates a comprehensive process for predicting real estate house prices using various machine learning techniques. The notebook walks through data exploration, preprocessing, model training, evaluation, and selection, ultimately culminating in a robust predictive model.

Table of Contents

• Introduction
• Dataset
• Installation
• Usage
• Notebook Overview
  • 1. Data Loading and Initial Exploration
  • 2. Exploratory Data Analysis (EDA)
  • 3. Data Cleaning and Preprocessing
  • 4. Model Training and Evaluation
  • 5. Hyperparameter Tuning
  • 6. Model Comparison
  • 7. Final Model and Pipeline
  • 8. Conclusion
• Contributing
• License

Introduction

This project aims to predict real estate hosue prices using a dataset containing various features such as area, furnishing status, and more. It employs multiple machine learning algorithms to find the best model for the task.

Dataset

The dataset used in this project is a structured CSV file containing information about real estate properties. Key features include price, area, and furnishing status.

Installation

The project uses Poetry to manage dependencies. To install the dependencies, run the following command:

poetry install

Notebook Overview

1. Data Loading and Initial Exploration

• Load the dataset.
• Display basic statistics and the first few rows of the data.
• Check for unique values and missing data.

2. Exploratory Data Analysis (EDA)

• Visualize the distribution of the target variable (price) using histograms and KDE plots.
• Examine the relationship between price and other features using scatter plots and box plots.
• Generate a heatmap to understand feature correlations.
• Create pair plots to visualize relationships between multiple features.

3. Data Cleaning and Preprocessing

• Handle binary categorical variables and create dummy variables for other categorical data.
• Standardize or normalize numerical features as needed.

4. Model Training and Evaluation

• Train multiple regression models including Linear Regression, Lasso, Ridge, Polynomial Regression, and Random Forest.
• Evaluate models using metrics like R², MSE, and RMSE.

5. Hyperparameter Tuning

• Use Optuna for hyperparameter tuning of the Random Forest model to improve performance.

6. Model Comparison

• Compare the performance of various models on the training and test sets.
• Visualize and interpret model performance metrics.

7. Final Model and Pipeline

• Save the best model using pickle.
• Create a prediction pipeline with the best model and necessary preprocessing steps.
• Demonstrate making predictions with the final model.

8. Conclusion

• Summarize findings, model performance, and insights gained from the project.