# Machine Learning House Price Prediction

An analysis of real-estate data that examines which factors influence house prices and compares regression models for predicting them.

- Role: Sole Developer
- Date: Jan 2025
- Categories: poc
- Tech: Python, NumPy, Pandas, Matplotlib, Seaborn, SciPy, Scikit-learn
- GitHub: https://github.com/AmulyaJain2004/HousePricePrediction

## Overview

This proof of concept studies a real-estate dataset in two steps: first understanding which factors influence house prices, then building machine-learning models that predict them. It follows the standard workflow for a tabular regression problem from raw data to a fitted model.

## System Design

```mermaid
flowchart TD
    Data[("Real estate dataset<br/>transaction date · house age · MRT distance<br/>convenience stores · latitude · longitude")] --> Clean["Data quality checks<br/>and cleaning"]
    Clean --> EDA["EDA<br/>distributions · correlations"]
    EDA --> FE["Feature engineering<br/>scaling · polynomial features · one-hot encoding"]
    FE --> Split["Train / test split"]
    Split --> Models["Linear · Ridge · Lasso<br/>Random Forest · Gradient Boosting"]
    Models --> Tune["Cross-validation +<br/>GridSearchCV"]
    Tune --> Eval(["MAE · MSE · R²"])
```

The target is house price per unit area, predicted from six features. Linear models (plain, Ridge and Lasso, with scaled and polynomial features built through scikit-learn pipelines) are compared with tree ensembles (Random Forest and Gradient Boosting). Hyperparameters are tuned with cross-validation and grid search, and the models are compared on mean absolute error, mean squared error and R².

## Analysis Components

- **Dataset overview and data quality assessment.**
- **Statistical summary and distribution analysis.**
- **Data cleaning and preprocessing.**
- **Exploratory data analysis (EDA).**
- **Feature engineering and correlation analysis.**
- **Model development** using regression techniques from scikit-learn.

## Status

A proof of concept for practising the end-to-end analysis workflow. It has no hosted demo.

---

Back to the full index: /agent