Personalized Context-aware Collaborative Online Activity Prediction
summary
Builds a user-location-time-activity 4D-tensor and a location-time-POI 3D-tensor, uses Telecom app logs (7 days, ~2000 apps) and Check-in data (3 months, 16 POI categories), and applies a transfer learning based collaborative tensor factorization to transfer time/location semantics from offline check-ins to online app prediction. With only 30% observed time-location contexts the method cuts sparsity effects and improves Top-5 prediction accuracy by about 40% (up to ~42–45% in reported settings) and raises nDCG5 by ~10%.
AI-written summary · about
chapters (6)
related talks
28m"Where, When, Why, and How: Directi
19mUsing the activity logs to predict user behaviour
15mAdding on-device recommendations to your app using TensorFlow and Firebase
06mPREFER: Point-of-interest REcommendation with efficiency and privacy-preservation via Federated ...
06mPerson-Centered Predictions of Psychological Constructs with Social Media Contextualized by ...
06mMetaTP: Traffic Prediction with Unevenly-Distributed Road Sensing Data via Fast Adaptation
12mTopology Over Predictions: Using Graph Context to Drive Real-Time Lending Decisions
21mCross-dataset Time Series Anomaly Detection for Cloud Systems