House Price Prediction System (Machine Learning + FastAPI)

Project Overview

This project is an end-to-end Machine Learning system that predicts house prices based on neighborhood and housing features such as crime rate, number of rooms, pollution level, and socio-economic factors.

The model is trained on the Boston Housing Dataset and deployed using FastAPI to provide real-time predictions through a REST API.

This project demonstrates practical ML engineering skills, not just model training.

---

Problem Statement

House prices depend on many factors:

• Crime rate
• Number of rooms
• Location advantages (river proximity)
• Pollution level
• Tax rate
• Socio-economic status

The goal is to learn patterns from historical data and predict the median house value (MEDV) accurately.

---

Dataset

• Source: Boston Housing Dataset
• Records: 506 houses
• Features: 13 numerical features
• Target: MEDV (Median value of owner-occupied homes)

Feature Description (Simple Meaning)

Feature	Meaning
CRIM	Crime rate in area
ZN	Residential land percentage
INDUS	Industrial area proportion
CHAS	Near Charles River (1 = Yes)
NOX	Air pollution level
RM	Average number of rooms
AGE	Age of houses
DIS	Distance to city centers
RAD	Road accessibility
TAX	Property tax rate
PTRATIO	Student-teacher ratio
B	Population demographic score
LSTAT	% of low-income population
MEDV	House price (Target)

---

Machine Learning Approach

• Problem Type: Regression

• Algorithm: Random Forest Regressor

• Why Random Forest?

  • Handles non-linear relationships well
  • Robust to outliers
  • Strong performance on tabular data

---

Model Performance

Metric	Value
MAE	~2.06
RMSE	~2.92
R² Score	~0.88

✔ Model explains ~88% of price variance
✔ Average prediction error ≈ ±5 price units

This is solid performance for this dataset.

---

Project Structure

house-price-prediction/
│
├── data/
│   └── boston.csv
│
├── notebooks/
│   └── eda_and_training.ipynb
│
├── model/
│   ├── house_price_model.pkl
│   └── scaler.pkl
│
├── main.py
├── requirements.txt
├── pyproject.toml
├── README.md
└── .gitignore

---

FastAPI Integration

The trained model is exposed via a REST API using FastAPI.

API Features

• Accepts house features as JSON
• Returns predicted house price
• Swagger UI available for testing

Run the API

fastapi run main.py

Open in browser:

http://127.0.0.1:8000/docs

---

Example API Input

{
  "CRIM": 0.3,
  "ZN": 12,
  "INDUS": 7.0,
  "CHAS": 0,
  "NOX": 0.47,
  "RM": 6.2,
  "AGE": 60,
  "DIS": 4.0,
  "RAD": 5,
  "TAX": 320,
  "PTRATIO": 16.5,
  "B": 380,
  "LSTAT": 14.0
}

Example Output

{
  "predicted_house_price": 23.84
}

Demo

---

Tech Stack

• Python
• Pandas, NumPy
• Scikit-learn
• Random Forest
• FastAPI
• Joblib

---

Key Learnings

• Real-world data preprocessing
• Feature importance analysis
• Regression evaluation metrics
• Model serialization
• API-based ML deployment
• Debugging ML pipelines (scaling issues)

---

Future Improvements

• Dockerize the application
• Deploy on cloud (Render / AWS / Railway)
• Add model versioning
• Add input validation & logging
• Try Gradient Boosting / XGBoost

---

Author

Ali Sulman
Aspiring Machine Learning Engineer
Focused on production-ready ML systems