The CHAMPS completed another busy week with the implementation of new features and further model improvements. The model engineering team improved training scripts for the sequence-to-sequence model, implemented specialized anomaly detection by event, and added dataset inferences. Meanwhile, the feature engineering team continued to fine-tune model effectiveness by completing features that monitor successive zero transactions, authorization number length, maximum pre-authorization sale amounts, pump numbers, and point of service numbers. Furthermore, based on team research and discussion of other feature analysis methods, the CHAMPS are exploring the efficiency of Linear Discriminant Analysis (LDA), Random Projections, and possibly Principal Component Analysis (PCA) as solutions for dimensionality reduction beyond Singular Value Decomposition (SVD). The CHAMPS are also preparing for the second Qualification Review Board (QRB) panel scheduled for next Tuesday. During the latest Thursday meeting, the team made improvements to the project presentation and deliverables based on the first QRB feedback. The upcoming blog will cover the team’s success in the second session of the Qualification Review Board.