House Price Prediction Project

This project focuses on predicting house prices based on various features of the property. We use a machine learning model to predict the SalePrice of properties.

Project Overview

The dataset used for this project contains multiple features related to the properties, including construction details, location, and quality attributes. Our goal is to predict the sale price (SalePrice) of a property.

Dataset Description

The dataset consists of various columns such as:

• SalePrice: The sale price of the property (target variable).
• MSSubClass: The construction class of the property.
• MSZoning: General zoning classification of the property.
• LotFrontage: The length of the property’s frontage on the street.
• LotArea: The size of the property (in square feet).
• Street: The type of street access.
• Alley: The type of alley access to the property.
• LotShape: The general shape and condition of the property.
• LandContour: The flatness or contour of the land.
• Neighborhood: The neighborhood where the property is located.
• OverallQual: The overall material and finish quality of the house.
• YearBuilt: The year the house was built.
• ... (Other features describe various aspects like foundation, basement, heating, and more).

Data Preprocessing & Feature Engineering

• Missing Value Handling: Missing values in several columns are handled, and some columns are filled with default values or median values.
• Outliers: Outliers in numerical variables are replaced with appropriate thresholds.
• Feature Engineering: Several new features are created, such as:
  • NEW_LAND_CONTOUR: Encodes the property’s contour.
  • NEW_YEAR_BUILT: Encodes different periods of house construction.
  • NEW_BSMT_VALUE: Combines multiple basement-related features into a single value.
  • NEW_FIREPLACES: Combines features related to fireplaces.

EDA (Exploratory Data Analysis)

During the exploratory data analysis (EDA), we examined the relationships between the target variable (SalePrice) and independent variables. Correlation analysis was performed, and irrelevant or highly correlated features were removed to improve model performance.

Models

Several machine learning models are used in this project, including:

• LightGBM Regressor
• Random Forest Regressor

The models are tuned using hyperparameter optimization techniques such as GridSearchCV.

Key Findings

• The SalePrice is highly correlated with features like OverallQual, GrLivArea, TotRmsAbvGrd, and others.
• Feature engineering plays a crucial role in improving model performance, particularly with the creation of new features based on domain knowledge.

Requirements

• Python 3.x
• Pandas
• Numpy
• Scikit-learn
• LightGBM

Installation

1. Clone the repository:

   git clone <repository_url>