noto

A dual input ML model for classifying Android notifications into 3 categories:

• critical - Payment alerts, OTPs, urgent security notifications
• important - Active navigation, food delivery tracking, appointment reminders
• not important - Promotional offers, media playback, social media updates

quick start

from src.classifier import DualInputNotificationClassifier
import pandas as pd

# load trained model
clf = DualInputNotificationClassifier.load('models/dual_classifier.joblib')

# predict
notifs = pd.DataFrame([{
    'app_label': 'GPay',
    'title': 'Payment received',
    'text': 'Rs 500 credited from Rahul'
}])

print(clf.predict(notifs))  # ['critical']

training

open notebooks/notification_classifier.ipynb and set TRAIN_MODEL = True.

model performance

• Accuracy: 94%
• CV Accuracy: 92.97% (±0.85%)
• Uses dual TF-IDF streams (app name + text) with RandomForest