DoluMu - Istanbul Public Transit Crowding Prediction Platform
ML-powered crowd and occupancy predictions for all Istanbul public transport. Check before you ride.
🌐 Available in Turkish and English | Türkçe ve İngilizce dillerinde kullanılabilir
What is this Platform?
DoluMu is a multilingual, web-based ML prediction tool that helps you avoid peak hours and plan more comfortable journeys on Istanbul's public transportation network covering Metro, Bus, Metrobus, and rail lines.
Important: This platform uses ML models trained on historical passenger data and weather forecasts to predict crowding levels. It does not rely on real-time sensors or live passenger counting. All information shown represents forecasted predictions based on patterns learned from past data, weather conditions, and calendar factors.
Key Features
🕐 Hourly Crowding Forecasts
- Hour-by-hour crowding for today and tomorrow on metro, Metrobüs, bus, Marmaray and ferry lines
- A "Right now" summary with a quieter-hour suggestion, so you can decide whether to wait
🌦️ Weather-Aware Predictions
- The model uses the weather forecast (rain, temperature, wind), since weather changes how people travel
- Current Istanbul weather is shown on the home screen
🚦 City Traffic Index
- Istanbul-wide congestion index from the İBB Traffic Management Center (UYM), refreshed every 5 minutes
🗺️ Network Map
- The home screen is a map of Istanbul's rail network and Metrobüs in official line colours
- Line width shows forecast riders for the chosen hour; scrub or play through the day
- Tap a line for its page, a station for the lines calling there; "stations near you" uses your location
- Line pages highlight the route: bus routes per direction, rail lines with every station
🚇 Rail Lines
- Metro, tram, funicular and cable-car lines use their official Metro İstanbul colours and names
- Operating hours from the Metro İstanbul topology; hours outside service are shown as gaps
M1 is searchable as its M1A / M1B branches; both share the same forecast
🚌 Departures & Notices
- First/last departure, next departures with countdown and the full daily timetable (planned İETT schedules)
- İETT service notices shown on the line page
📱 Mobile-First Web App
- One responsive design: single column on phones, two columns on large screens
- Installable to the home screen (Add to Home Screen / Install app)
- Light, dark or system theme; shareable line URLs (
/tr/line/500T?dir=D&day=tomorrow)
🌐 Multi-Language Support
- Full interface available in Turkish (Türkçe) and English
- Easy language switching from Settings page
- All forecasts, labels, and notifications localized
- Automatic language detection based on browser preferences
How to Read the Crowd Levels
Levels are relative to each line's own busiest hour of that day, like "Popular times" on Google Maps:
| Level | Share of the day's busiest hour | Meaning |
|---|
| 🟢 Quiet | < 40% | One of the quietest times of the day |
| 🟡 Normal | 40–65% | Typical for this line |
| 🟠 Busy | 65–85% | Crowded, seats may be hard to find |
| 🔴 Very busy | ≥ 85% | Among the busiest hours of the day |
The "Right now" card also suggests a noticeably quieter hour within the next three hours when there is one.
Capacity-based occupancy (occupancy_pct) is still shown under "How is this forecast made?", together with a note on how reliable the capacity estimate is. It is not used for the colours because the capacity estimates are coarse and saturate at 100% on many lines.
🚦 Service Awareness
- Hours that fall outside the published schedule are clearly marked Out of Service, so you instantly know when a line is offline instead of staring at an empty chart.
- Direction-aware status checks mean the app can tell you if only one side of a route (e.g.,
G vs D) is paused, keeping the forecast, status banner, and schedule widget perfectly aligned.
- Metro/rail out-of-service hours are derived from Metro topology service windows (
first_time/last_time) so 24h charts remain visible while inactive hours render as gaps.
User Scenarios
🏢 Daily Commuters
"I work in Levent and live in Kadıköy. Should I leave at 17:30 or wait until 18:30?"
- Check M2 Metro predictions for both time slots
- Compare crowd levels and choose the more comfortable option
🛍️ Weekend Shoppers
"Planning to visit Taksim on Saturday afternoon - when will the metro be less crowded?"
- View weekend patterns for M2 Vezneciler → Taksim
- Get suggestions for off-peak shopping hours
🏥 Medical Appointments
"I have a doctor's appointment in Bakırköy at 14:00 - when should I leave Beylikdüzü?"
- Check Metrobüs predictions 2-3 hours before departure
- Plan buffer time based on predicted crowding levels
🎯 Event Attendees
"There's a match at Vodafone Park tonight - how crowded will the metro be?"
- Check predictions for lines serving Beşiktaş
- Plan alternative routes if main lines show high crowding
FAQ & Important Information
How accurate are these predictions?
Our ML models achieve good accuracy for typical conditions, but predictions are estimates, not guarantees. Accuracy is highest for:
- Regular weekday patterns
- Well-established metro and metrobüs lines
- Normal weather conditions
Predictions may be less accurate during:
- Unusual events (strikes, major celebrations, emergencies)
- Extreme weather conditions not seen in historical data
- New transportation routes with limited historical data
This is the beginning of the README. Read the rest at the original. Read the full README ↗
