Our Capstone project Client, GE Aviation has a product called Ceramic Matrix Composite (CMC). CMC products are used in Jet Engines and they can have variable build patterns. These build patterns are stored as Bill of Materials (BOM). The GE team required human expertise to detect anomalies in this Bill of Materials. This consumed a lot of company resources. The purpose of this project was to prototype data mining-based methodology for forming generic traceability pattern from legacy databases and build mechanism to alert the GE team for outliers.

I, with the help of my two other teammates, came up with the following solutions to prototype a data mining-based methodology for forming generic traceability pattern from legacy databases and build mechanism to alert the GE team for outliers:

1. Used python's networkX library to parse the Bill of Materials and produce appropriate graph structures for creating the feature space for anomaly detection.

2. Used DBSCAN implemented in scikit-learn library to form clusters that will separate the normal traces from the anomalous ones.

3. Used pandas library to load, store and manipulate data.

4. Identified two categories of anomalies within the dataset given to us.