As the CHAMPS settled back into their normal school routines after a successful Preliminary Design Review (PDR) presentation last week, the necessary project revisions loomed ahead. Week 9 required the team to implement the adjustments into the presentation and report within a couple of business days post-visit. Since the project pivoted due to certain data restrictions, the CHAMPS are focusing on one type of log from the Commander going forward. The brainstorming session with Verifone from the previous week also clarified that the initial “health analysis” intent of the project was not feasible given the data the Commander currently collects. Instead, the CHAMPS will build a machine-learning model that identifies anomalous events such as potential fraud and provides warnings accordingly to merchants. The team will continue feature selection on the data available since the liaison engineers are encountering problems receiving more data necessary to train the future model.

However, now that the PDR presentation and report are updated according to the Verifone site visit meeting, team members are tasked with a variety of data analyses and research. The CHAMPS are exploring the use of recurrent neural networks (RNN) and investigating methods to scale the model for the amount of data expected from a live Commander. The team is also preparing the data analysis and preliminary model for Prototype Inspection Day on November 14th which the blog will report more on next week.