🔔 Notification Bad-Timing Detector

Should you send that push notification right now? This model predicts the probability that the current moment is a bad time to interrupt the user, based on their activity patterns, battery status, and notification interaction history.

Built with LightGBM + isotonic calibration • Trained on 100K samples • Model • Dataset

⚡ Quick presets — try a scenario

⏰ Time Context

Hour of day (0–23)

0 23

Day of week

🔋 Battery Status

Battery level (%)

0 100

Charging?

Battery drain rate (%/hr)

-5 5

📱 User Activity

Screen on?

Screen-on time last 30 min (sec)

0 1800

App opens last hour

0 30

Current session length (sec)

0 3600

Time since last interaction (sec)

0 14400

🔔 Notification History

Shown (30 min)

0 20

Clicked (30 min)

0 20

Dismissed (30 min)

0 20

Ignored (30 min)

0 20

Shown (24h)

0 100

7-day CTR

0 1

Recent notification density

0 10

📊 Decision Thresholds

Probability	Action
< 30%	✅ Send notification
30–50%	⚠️ Only if important
50–80%	🚫 Delay notification
> 80%	🔴 Definitely delay

🧠 How it works

The model uses 21 signals across 4 categories:

Time: hour, day, weekend/night flags
Battery: level, charging, drain rate
Activity: screen, app opens, session length
Notifications: recent shown/clicked/dismissed/ignored, 7-day CTR

Built on research from the C-3PO paper (Cheetah Mobile, 600M monthly active users).