Utilizing Word Embeddings and Gradient Boosting
summary
Combines log word embeddings with gradient boosting on 23 sensor columns to consolidate unstructured logs and tabular streams from dental milling machines into a single feature set, builds an end-to-end ML pipeline that clusters embedding vectors to surface recurring error modes, uses embedding features inside tree models for prediction and feature importance, addresses class imbalance, and reports a roughly 10x accuracy boost in early experiments for error prognosis and prevention.
AI-written summary · about
chapters (6)
- 0:00Error prediction in milling
- 1:33Prognosis prevention for errors
- 3:42Data challenges and boosting
- 5:30Word embeddings for logs
- 7:15Embeddings with boosting
- 8:42Real-world impact and performance
related talks
24mHow to numerically represent semi-structured log data for...
31mTaking AI/ML to the next level: Text!
10mML-based Log Analysis for Faster Debugging
24mHybrid ML Methods for the Interpretation and Integration of Heterogeneous Multimodal Data
43mDemocratizing NLP content modeling with transfer learning using GPUs
17mAnnotating Log Data For Machine Learning
01h 08mGoing multi-modal
16mMachine Learning on Source Code