Ames Housing Price Prediction 🏠📈

This repository contains an end-to-end regression project using the
Ames Housing dataset from Kaggle.

The goal is to predict house sale prices by applying:

• classical data preprocessing techniques
• feature engineering
• and a Neural Network built using TensorFlow Keras Functional API

This project is primarily for learning and experimentation, not leaderboard chasing.

---

📌 Dataset

• Source: Kaggle Competition
  House Prices: Advanced Regression Techniques
• Target variable: SalePrice
• Data includes:
  • numerical features
  • categorical features
  • missing values
  • skewed distributions

⚠️ Dataset files are not included in this repository due to Kaggle’s licensing rules.

---

🗂️ Project Structure

.
├── data/ # Local dataset (ignored by git)
├── notebooks/ # EDA & experiments
├── src/ # Preprocessing & model code
├── models/ # Saved models (optional)
├── README.md
├── .gitignore
└── requirements.txt

---

🔁 Workflow / Steps Followed

1️⃣ Data Acquisition

• Download dataset using Kaggle CLI
• Store raw data locally under data/

---

2️⃣ Exploratory Data Analysis (EDA)

• Understand feature distributions
• Analyze target variable (SalePrice)
• Identify:
  • missing values
  • skewness
  • outliers
• Visualizations:
  • histograms
  • correlation heatmaps
  • pairplots (feature vs target)

---

3️⃣ Data Preprocessing

• Handle missing values
• Encode categorical variables
• Scale numerical features
• Log-transform skewed variables
• Train–test split

---

4️⃣ Model Building (Keras Functional API)

• Define input layers explicitly
• Build dense neural network with:
  • multiple hidden layers
  • ReLU activations
  • output layer for regression
• Compile model with:
  • optimizer (Adam)
  • loss function (MSE / MAE)

---

5️⃣ Model Training

• Train on training data
• Validate on hold-out set
• Monitor:
  • training loss
  • validation loss
• Tune:
  • number of layers
  • neurons
  • learning rate

---

6️⃣ Evaluation

• Evaluate model using:
  • RMSE
  • MAE
• Compare with baseline ML models (optional)

---

7️⃣ Experimentation & Improvements

• Feature selection
• Regularization (Dropout / L2)
• Hyperparameter tuning
• Architecture experiments

---

🧠 Why Functional API?

• More flexible than Sequential
• Explicit control over:
  • inputs
  • outputs
  • complex architectures
• Industry-standard for non-trivial models

---

🛠️ Tech Stack

• Python
• NumPy, Pandas
• Matplotlib, Seaborn
• Scikit-learn
• TensorFlow / Keras
• Kaggle API
• Git & GitHub

---

🚫 Notes

• data/ directory is ignored using .gitignore
• Kaggle datasets should not be committed
• This repo focuses on learning and clarity, not competition ranking

---

📜 License

This project is licensed under the Apache License 2.0
See the LICENSE file for details.

---

✍️ Author

Ashiq KM
Learning-focused ML & Deep Learning experiments 🚀