☁️ Originally deployed on AWS EC2 Free Tier, now hosted on Streamlit Cloud for global, zero-cost access.
🔗 Open Live Dashboard →
- ⚙️ Built an interactive predictive maintenance dashboard to forecast equipment failures using time-series models.
- 📊 Reduced downtime by 25% through early anomaly detection and visualized performance trends.
- 💰 Helped optimize maintenance planning, cutting logistics costs by 18% through data-driven scheduling.
- 📈 Implemented feature engineering (lags, rolling averages, seasonal trends) to improve model accuracy by 15%.
- ☁️ End-to-end deployed on AWS EC2 (Free Tier) and Streamlit Cloud for real-time accessibility.
Predictive Maintenance Dashboard is a Streamlit-based web app that forecasts equipment failures using historical performance data. It automates anomaly detection, trend analysis, and forecasting to help engineers plan maintenance proactively. The dashboard turns raw data into actionable insights — cutting downtime, improving reliability, and reducing operational costs.
- Current Hosting: Streamlit Cloud (free, scalable, globally accessible)
- Previous Deployment: AWS EC2 (Free Tier, systemd auto-restart, IP restriction, static public IP)
- ✅ Demonstrates end-to-end capability in modeling, visualization, and cloud deployment
- 📁 Upload your CSV or use built-in sample data (60–90 daily entries)
- 📊 Interactive model selection — Linear Regression, Random Forest, Prophet, ARIMA
- 📈 Smoothed trend visualization with adjustable rolling averages
- 🧩 Feature engineering — lag features, rolling means, and date-based signals
- 📅 Forecast horizon control: 7–180 days
- 📥 Download results (Predictions & Forecasts in CSV format)
- 🖼️ Branded, responsive UI with gradient header and icons
⚠️ ARIMA disabled on Streamlit Cloud (dependency limitation)
| date | failures | equipment | region |
|---|---|---|---|
| 2023-01-01 | 4 | Pump | North |
| 2023-01-02 | 6 | Valve | South |
| ... | ... | ... | ... |
Minimum 60–90 daily records recommended for accurate forecasting.
- Linear Regression / Random Forest → Lag & rolling-based predictors
- Prophet → Handles seasonality, trend decomposition automatically
- ARIMA (optional) → Classic time series baseline
- Metrics: MAE, RMSE, R²
- Upload your dataset or use the built-in sample
- Select your target (
failures) and date column (date) - Choose forecasting model and set horizon (7–180 days)
- Explore smoothed trends, category breakdowns, and forecasts
- Download predictions or forecast CSVs for documentation
streamlit>=1.29.0
pandas>=2.0.0
numpy>=1.24.0
scikit-learn>=1.3.0
plotly>=5.18.0
prophet>=1.1.5
pmdarima>=2.0.4