🚀 End-to-end machine learning system that predicts potential failures using operational data and provides real-time alerts via an interactive dashboard.
- 🔍 Detects potential system failures using ML
⚠️ Real-time alert system (Failure / Normal prediction)- 📊 Handles real-world issues like data leakage & class imbalance
- 🖥️ Deployed with an interactive Streamlit dashboard
- 📈 Achieves ~70% accuracy with improved failure detection
This project simulates an AI-based predictive maintenance system used in industries to prevent unexpected failures.
It includes:
- Data preprocessing
- Feature engineering
- Machine learning model (Random Forest)
- Real-time prediction dashboard
Unexpected failures in industrial systems lead to:
- Downtime
- Financial loss
- Reduced efficiency
This system predicts failures in advance using data-driven techniques.
- Manufacturing plants
- Automotive systems
- IoT-based monitoring
- Energy & power systems
- Python
- Pandas, NumPy
- Scikit-learn
- Matplotlib
- Streamlit
- Walmart dataset used to simulate operational conditions
- Failure defined using proxy-based labeling (low performance)
-
Algorithm: Random Forest Classifier
-
Improvements:
- Removed data leakage
- Handled class imbalance using weighting
- Accuracy: ~70%
- Improved recall for failure detection
Data → Preprocessing → Feature Engineering → Model → Prediction → Dashboard
git clone <your-repo-link>
cd AI-Predictive-Maintenance-System
pip install -r requirements.txt
python main.py
streamlit run app.py- Handling real-world data challenges
- Avoiding data leakage
- Managing imbalanced datasets
- Building end-to-end ML systems
- Deploying ML models using Streamlit
- Real IoT sensor integration
- Deep learning models (LSTM)
- Real-time streaming pipelines
- Cloud deployment
Varda Kunde