IoT-Paper: Machine Learning for IoT / Windows Dataset Classification

A clean, reproducible machine-learning project for experimenting with IoT/Windows dataset classification.
The original repository contained a Jupyter notebook and a CSV dataset. This upgraded version reorganizes the work into a GitHub-ready research project with a professional README, source code modules, configuration files, reproducible scripts, and documentation.

Research note: The original notebook reported very high scores with models such as LightGBM and Decision Tree. This version adds a safer evaluation pipeline to reduce the risk of data leakage and overfitting.

---

✨ Features

• Clean project structure for GitHub and research presentation
• Reproducible ML pipeline using scikit-learn and lightgbm
• Binary and multiclass classification support
• Data validation, preprocessing, model training, evaluation, and report generation
• Confusion matrix and classification report outputs
• Configuration-based workflow
• GitHub Actions workflow for basic Python checks
• Ready-to-customize notebook folder

---

📁 Project Structure

IoT-Paper-Professional/
├── configs/
│   └── default.yaml
├── data/
│   ├── raw/              # Put windows7_dataset.csv here
│   └── processed/        # Generated clean datasets
├── notebooks/
│   └── README.md
├── reports/
│   └── figures/          # Saved plots and evaluation figures
├── src/
│   └── iot_paper/
│       ├── __init__.py
│       ├── data.py
│       ├── features.py
│       ├── train.py
│       └── evaluate.py
├── .github/workflows/
│   └── python-checks.yml
├── .gitignore
├── CITATION.cff
├── CONTRIBUTING.md
├── LICENSE
├── pyproject.toml
├── requirements.txt
└── README.md

---

🚀 Quick Start

1. Clone the repository

git clone https://github.com/mohammadali-mousavireineh/IoT-Paper.git
cd IoT-Paper

2. Create a virtual environment

python -m venv .venv

Windows:

.venv\Scripts\activate

macOS / Linux:

source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Put the dataset in the correct folder

Place your dataset here:

data/raw/windows7_dataset.csv

5. Run training

python -m iot_paper.train --config configs/default.yaml

---

🧠 Machine Learning Pipeline

The pipeline follows these steps:

1. Load raw CSV data
2. Detect and remove empty / duplicated columns
3. Split features and target label
4. Encode categorical values
5. Train a baseline model
6. Evaluate with accuracy, F1-score, precision, recall, and confusion matrix
7. Save results under reports/

---

⚠️ Important Research Warning: Overfitting & Leakage

Very high accuracy, such as 99%, can happen for two very different reasons:

1. The model is genuinely good, or
2. The dataset leaks the answer through duplicated rows, target-like columns, timestamps, IDs, or train/test contamination.

Before claiming strong performance, check:

• Are there duplicated rows across train/test sets?
• Are there columns that directly reveal the target?
• Is the split random when it should be time-based or device-based?
• Are IDs, session labels, or post-event columns included as features?
• Does cross-validation show similar performance?

---

📊 Suggested Experiments

• Decision Tree baseline
• Random Forest baseline
• LightGBM multiclass classifier
• Binary classification version
• Feature importance analysis
• Ablation study: remove suspicious columns and retrain
• Stratified K-fold cross-validation
• Time-aware or group-aware split if session/device IDs exist

---

🧪 Example Commands

Train using the default configuration:

python -m iot_paper.train --config configs/default.yaml

Run evaluation only after saving predictions:

python -m iot_paper.evaluate --predictions reports/predictions.csv

---

📌 Dataset

The dataset is not redistributed in this professional template. Put the original CSV file in:

data/raw/windows7_dataset.csv

The original GitHub repository includes windows7_dataset.csv and Untitled-1.ipynb.

---

🛣️ Roadmap

• Clean and rename the original notebook
• Add exploratory data analysis notebook
• Add leakage detection report
• Add model comparison table
• Add feature importance visualizations
• Add paper-style result summary
• Add Dockerfile
• Add Streamlit demo dashboard

---

👤 Author

Mohammadali Mousavireineh
GitHub: @mohammadali-mousavireineh

---

📄 License

This project is released under the MIT License. See LICENSE for details.