Graduate mentor's supervisor: Grant Weddell
Research Problem
Understanding how datasets in a massive data lake relate to one another is incredibly difficult because documentation is usually outdated. Instead of looking at static, broken files, this research takes a radical approach: it discovers data relationships by analyzing how people actually use the data. Solving this reduces manual data cleaning and helps organizations automatically map out how their data flows and changes in real time.
Importance:
You get to build a Behavioral Knowledge Graph by studying query logs and user histories. Instead of guessing what the data means from a rigid schema, you are teaching a system to discover hidden relationships dynamically based on real human behavior. It shifts the focus from data as stored to data as used!
As a team of 3–4, you will manage balanced weekly tasks such as:
- Short-term: Parse database query logs to see which tables are frequently used together.
- Medium-term: Write algorithms to automatically build a behavioral graph from that log data.
- Longer-term: Build a recommendation tool that suggests useful datasets to users based on team habits.