raj8668/

Engine-Predictive-Maintenence

Predictive Maintenance – Engine Failure Classification (MLOps Project)

🎉 Status: Complete - All pipeline steps successfully deployed!

• ✅ Data registration on Hugging Face
• ✅ Exploratory Data Analysis (EDA)
• ✅ Data preparation and dataset versioning
• ✅ Model building with experimentation tracking
• ✅ Model deployment with Docker + Streamlit on Hugging Face Spaces
• ✅ Automated GitHub Actions workflow

🔗 Live Resources

🌐 Live Application

🤖 Trained Model

📊 Dataset Repository

💻 GitHub Repository

⚙️ GitHub Actions

📁 Repository Structure

mlops/
├── data/                          # Raw and processed data
│   ├── engine_data.csv           # Original engine sensor dataset
│   └── processed/                # Train/test splits
│       ├── train.csv
│       └── test.csv
├── notebooks/                     # EDA and experimentation notebooks
├── src/                          # Main source code
│   ├── config.py                 # Central configuration
│   ├── data_register.py          # Register raw data to HF Dataset
│   ├── data_prep.py              # Data cleaning and splitting
│   ├── hf_data_utils.py          # HF Dataset Hub utilities
│   ├── train.py                  # Model training with MLflow
│   ├── hf_model_utils.py         # HF Model Hub utilities
│   ├── inference.py              # Prediction utilities
│   ├── app.py                    # Streamlit web application
│   └── deploy_to_hf.py          # Deploy to HF Space
├── .github/
│   └── workflows/
│       └── pipeline.yml           # CI/CD pipeline
├── Dockerfile                    # Container definition for deployment
├── requirements.txt              # Python dependencies
└── README.md                     # This file

Key Files

• src/config.py – Central configuration (paths, Hugging Face repo names, MLflow config)
• src/data_register.py – Registers raw dataset to Hugging Face Dataset Hub
• src/data_prep.py – Loads data, cleans it, and creates train/test splits
• src/train.py – Model training, hyperparameter tuning, MLflow logging
• src/app.py – Streamlit web application for interactive predictions
• src/deploy_to_hf.py – Deploys app to Hugging Face Space
• .github/workflows/pipeline.yml – Automated CI/CD pipeline

🔄 Pipeline Overview

1. Data Registration ✅

• Script: src/data_register.py
• Action: Creates/uses Hugging Face dataset repo and uploads raw data
• Output: ananttripathiak/engine-maintenance-dataset with data/engine_data.csv

2. Exploratory Data Analysis (EDA)

• Script: src/eda.py (or use notebooks)
• Action: Performs data overview, univariate/bivariate/multivariate analysis
• Output: Visualizations and insights about engine health patterns

3. Data Preparation ✅

• Script: src/data_prep.py
• Action: Cleans data, creates train/test splits, uploads to dataset repo
• Output: data/train.csv and data/test.csv in dataset repo

4. Model Building + Experiment Tracking ✅

• Script: src/train.py
• Action: Trains Random Forest with hyperparameter tuning, logs to MLflow, uploads best model
• Output: ananttripathiak/engine-maintenance-model with trained model

5. Deployment & Hosting ✅

• App: src/app.py - Streamlit web application
• Container: Dockerfile - Container definition
• Script: src/deploy_to_hf.py - Deploys to Hugging Face Space
• Output: Live app at ananttripathiak/engine-maintenance-space

6. GitHub Actions Workflow ✅

• File: .github/workflows/pipeline.yml
• Jobs:
  1. register-dataset → runs src/data_register.py
  2. data-prep → runs src/data_prep.py
  3. model-training → runs src/train.py
  4. deploy-hosting → runs src/deploy_to_hf.py
• View Runs: GitHub Actions

🚀 Quick Start

Local Development

1. Clone the repository:
   bash

   1git clone https://github.com/ananttripathi/engine-predictive-maintenance.git 2cd engine-predictive-maintenance
2. Set up virtual environment:
   bash

   1python -m venv .venv 2source .venv/bin/activate # On Windows: .venv\Scripts\activate 3pip install -r requirements.txt
3. Run pipeline steps:
   bash

   1# Register data 2python src/data_register.py 3 4# Prepare data 5python src/data_prep.py 6 7# Train model 8python src/train.py 9 10# Run app locally 11streamlit run src/app.py

Automated Pipeline

📊 Technologies Used

• Python 3.10 - Programming language
• scikit-learn - Machine learning (Random Forest)
• MLflow - Experiment tracking and model registry
• Hugging Face Hub - Dataset, model, and space hosting
• Streamlit - Web application framework
• Docker - Containerization
• GitHub Actions - CI/CD automation
• Plotly - Interactive visualizations

⚙️ Configuration

Current Configuration

• Hugging Face Username: ananttripathiak
• GitHub Username: ananttripathi
• Dataset Repo: ananttripathiak/engine-maintenance-dataset
• Model Repo: ananttripathiak/engine-maintenance-model
• Space Repo: ananttripathiak/engine-maintenance-space
• GitHub Repo: ananttripathi/engine-predictive-maintenance

For New Users: Setup Instructions

1. Hugging Face Configuration

2. GitHub Repository Configuration

1. Create a new repository on GitHub (e.g., engine-predictive-maintenance)
2. Push this mlops folder to it:
   bash

   1git init 2git add . 3git commit -m "Initial commit: Predictive maintenance MLOps pipeline" 4git remote add origin https://github.com/your-username/engine-predictive-maintenance.git 5git push -u origin main

• HF_TOKEN – Your Hugging Face access token (from https://huggingface.co/settings/tokens)
• HF_DATASET_REPO – e.g., ananttripathiak/engine-maintenance-dataset
• HF_MODEL_REPO – e.g., ananttripathiak/engine-maintenance-model
• HF_SPACE_REPO – e.g., ananttripathiak/engine-maintenance-space