🤖 ML Stock Movement Classifier with MLOps

Welcome to ML_Stock_MLOps, a fully automated and production-ready MLOps project designed to classify stock movement trends using historical Amazon stock data. This repository demonstrates how to set up a robust ML pipeline with:

• ⚙️ GitHub Actions for CI/CD
• 📦 DVC (Data Version Control) for dataset + model tracking
• 📊 Weights & Biases (W&B) for experiment logging
• 🧠 Scikit-Learn for model training
• ☁️ Google Drive for remote storage

---

📁 Project Structure

ML_Stock_MLOps/
├── .dvc/                    # DVC metadata
├── .github/
│   └── workflows/
│       └── run-model.yml   # GitHub Actions workflow
├── fetch_data.py           # Dataset download script
├── model.py                # Training + logging
├── data.csv.dvc            # DVC pointer to dataset
├── dvc-sa-key.json         # (Ignored) GDrive auth key
├── model.pkl               # Saved model artifact
├── requirements.txt        # Dependencies
└── README.md               # This file

---

🚀 Features

🔁 1. GitHub Actions CI/CD

Every push to main branch automatically:

• Pulls the dataset from Google Drive
• Trains the model using model.py
• Logs metrics to W&B (accuracy)
• Uploads model.pkl as an artifact

📦 2. Data Versioning with DVC

• DVC tracks data.csv via data.csv.dvc
• The actual file is stored on Google Drive
• Enables consistent, reproducible experiments

📊 3. Experiment Tracking with W&B

• Accuracy and metadata are logged per run
• Visualize comparisons across different commits

---

🧪 Example Metrics Logged

Run	Accuracy	Model Version	Notes
#1	0.812	v1	Initial baseline
#2	0.835	v2	Tuned hyperparams

🔗 View logs on: W&B Dashboard

---

📦 Installation and Setup

1. Clone the Repository

git clone https://github.com/Hasee10/ML_Stock_MLOps.git
cd ML_Stock_MLOps

2. Install Dependencies

pip install -r requirements.txt

3. Setup DVC Remote with GDrive

dvc remote add -d gdrive_remote gdrive://<your-folder-id>
dvc remote modify gdrive_remote gdrive_use_service_account true
dvc remote modify gdrive_remote gdrive_service_account_json_file_path dvc-sa-key.json

4. Pull the Dataset

dvc pull

5. Train the Model Locally

python model.py

---

🔄 Workflow Automation

GitHub Actions

The workflow in .github/workflows/run-model.yml:

• Runs on every push to main
• Pulls data using DVC
• Trains the model
• Logs to W&B
• Uploads model.pkl

No manual intervention needed after setup! ✅

---

📤 Deployment Ideas (Next Steps)

Step	Tools You Can Use
Live Inference	Streamlit / Flask
Model Registry	MLflow / W&B Artifacts
Auto Retraining	Cron + GitHub Actions
Data Drift Monitor	Evidently / Grafana

Let me know if you’d like to implement these — happy to guide you!

---

💡 Why This Project Stands Out

✔️ Real-world data from Amazon stock history
✔️ End-to-end pipeline from data to deployment
✔️ Resume-ready, client-attracting project
✔️ Showcases production-level MLOps principles

---

👨‍💻 Author

Muhammad Haseeb Arshad
📧 haseeb.arshad@example.com
🔗 LinkedIn

---

⭐ Star This Repo If You Like It!

Your support keeps the projects coming 🚀

---

Built with ❤️ using Python, GitHub Actions, DVC, and W&B.