Using the activity logs to predict user behaviour
summary
Claims a working prototype that predicts each user’s next page from activity logs and uses the forecast to preload data and estimate server load. Model uses TensorFlow and TensorFlowJS in-browser inference, transforms pages with one-hot encoding, and trains a sequential dense model with ReLU and softmax (example run uses 100 epochs). Prototype runs tens of models per app instance, performs real-time predict calls to decide prefetches, but provides no rigorous accuracy numbers beyond illustrative examples.
AI-written summary · about
chapters (8)
- 0:00Introduction and Overview
- 0:37Understanding AI and Machine Learning
- 1:07How Machine Learning Models Work
- 3:08Choosing the Right Machine Learning Model
- 4:10Implementing Neural Networks
- 7:21Practical Application with TensorFlow
- 13:17Training and Using the Model
- 18:31Conclusion and Q&A
related talks
09mBuild a UI that Learns - Intelligent Prefetching
34mPredictive preloads for ultimate web performance?
22mVue + TensorFlow.js: Building Your First ML Powered App
31mA Feasibility of On-Device ML for Adaptive User Interfaces in Frontend Development
16mJavaScript Text Analytics for User Behavior Intelligence
32mUsing Machine Learning Models in JavaScript Application
48mDeep Learning in the Browser
29mAdvanced UX with tensorflow.js