| title: 🔔 Notification Bad-Timing Detector | |
| emoji: 🔔 | |
| colorFrom: yellow | |
| colorTo: red | |
| sdk: gradio | |
| sdk_version: 6.13.0 | |
| app_file: app.py | |
| pinned: false | |
| models: | |
| - alianassmaaa/notification-bad-timing-detector | |
| datasets: | |
| - alianassmaaa/notification-timing-dataset | |
| tags: | |
| - notification-timing | |
| - interruptibility | |
| - lightgbm | |
| - tabular-classification | |
| short_description: Should you send that push notification right now? | |
| # 🔔 Notification Bad-Timing Detector — Interactive Demo | |
| **Should you send that push notification right now?** | |
| This Space provides an interactive demo for predicting the probability that the current moment is a **bad time** to interrupt a user with a push notification. | |
| ## Features | |
| - 🎛️ **21 input signals** across 4 categories: time, battery, activity, notifications | |
| - ⚡ **5 preset scenarios** to quickly test common situations | |
| - 📊 **Visual probability gauge** with color-coded recommendations | |
| - 🎯 **Calibrated probabilities** via isotonic regression | |
| ## Model | |
| - **Architecture**: LightGBM gradient boosted trees with isotonic calibration | |
| - **Performance**: ROC-AUC 0.834, F1 0.777, Accuracy 75.8% | |
| - **Training**: 100K synthetic samples based on [C-3PO](https://arxiv.org/abs/1803.00458) research | |
| ## Links | |
| - 📦 [Model on HF Hub](https://huggingface.co/alianassmaaa/notification-bad-timing-detector) | |
| - 📊 [Dataset on HF Hub](https://huggingface.co/datasets/alianassmaaa/notification-timing-dataset) | |