https://meltdown-pattern-predictor-adahacks.vercel.app/
A smart tracking tool that helps caregivers of autistic children identify what triggers meltdowns—so families can plan better, stress less, and support their kids with confidence.
Built at Ada Hacks.
Caregivers of autistic children often feel like they're navigating meltdowns in the dark. A difficult day happens, and in the moment it's hard to pinpoint why—Was it the lack of sleep? The loud environment? The change in routine? The late-night screen time?
Without data, it's guesswork. Families replay events in their heads, wonder what they could have done differently, and hope to spot patterns before the next meltdown. But human memory is unreliable, and subtle triggers add up in ways that are easy to miss.
The result: Exhausted caregivers, repeated meltdowns, and missed opportunities to create calmer, more predictable environments that support autistic children's well-being.
- Rumaiza Norova
- Aruzhan Kapsemetova
Meltdown Pattern Predictor turns guesswork into insight. Caregivers log simple daily factors—sleep, sensory environment, routine changes, late meals, screen time—and whether a meltdown occurred. Our tool analyzes the data and surfaces which conditions correlate most strongly with meltdowns.
- 📝 Daily Logging – Quick, structured entries for sleep hours, sensory environment (noise level), late sugar, late screens, routine changes, late meals, and meltdown occurrence
- 🔮 Risk Prediction – XGBoost model estimates meltdown likelihood from today's factors. Shows which risk factors matter most and a bar chart of how much each would help if improved. What if? buttons let you explore scenarios (e.g., "Sleep 8h: 30%") before they happen.
- 📊 Pattern Analysis – After a few logs, the app computes meltdown rates across each factor and ranks the most likely triggers from your data
- 🧠 Data-Driven Insights – No more relying on memory. See exactly which conditions preceded meltdowns
- 🔒 Privacy-First – Everything stays in your browser. No accounts, no servers, no data leaving your device
- ✅ Actionable – Use the insights and What if scenarios to plan routines, avoid known triggers, and create more supportive environments
Parents, guardians, and caregivers of children on the autism spectrum who want to understand what contributes to meltdowns—not to blame, but to build better days.
Local:
- Clone or download this repo
- For AI help: Add your OpenAI API key in
config.js - Open
index.htmlin a web browser - Add daily logs (sample data loads by default so you can try it right away)
- Fill in today's factors and click Predict Risk for an estimated meltdown likelihood, risk factors, and an impact chart. Use the What if? buttons to see how improving one factor changes the prediction
- Click Analyze Patterns to view trigger insights from your saved logs
- Click the Help button (bottom-right) for AI assistance
Deploy on Vercel:
- Push to GitHub and import the repo in Vercel
- Add OPENAI_API_KEY in Project Settings → Environment Variables
- Deploy — the AI chatbot will use the serverless API and your env var
- HTML / CSS / JavaScript – Vanilla front-end
- Chart.js – Visualizations
- XGBoost – Meltdown risk prediction (precomputed grid, no server needed)
- OpenAI API – AI help chatbot (gpt-4o-mini)
- localStorage – Client-side data persistence
The app uses a precomputed predict_grid.json from a trained XGBoost model. To regenerate it (e.g., after changing synthetic data or model parameters):
cd synthetic-data-generator && python3 generate.py --count 10000
cd ../model-training && python3 train.py && python3 export_grid.pyThis produces a new predict_grid.json in the project root. Commit it to update the deployed app.
This tool does not diagnose autism or predict behavior with certainty. It is designed to help caregivers notice patterns that may support better routine planning and understanding. Always consult healthcare providers for medical or behavioral advice.