Most transactions, or collisions, show no anomalies. But when they do, this may lead to new ground-breaking insights for both economists and physicists.
>Hence this new collaboration plans to combat fluctuations in markets caused by anomalies, by combining the unique commodity and financial market data and understanding from CORMEC and WUR with CERN’s ROOT data analysis expertise and techniques.
Forgive me for snickering.
The story isn't about any breakthrough--just the announcement of a research partnership. Maybe they'll come up with something useful, but I'd bet on more false positives than detections of nefarious activity. (How do you detect insider trading?)
ROOT is a kitchen-sink tool. It allows you to structure and manipulate your data, manage I/O, and provides huge number of tools for fitting, analysis, plotting, and various other goodies as well. About a decade ago they did a static analysis of the code and found all sorts of problems that I hope they've fixed. If you restricted yourself to the high level routines you were pretty much OK, but if you had something more complex (e.g. an event display, or low level I/O), it was an uphill climb to migrate to new versions of ROOT. Been there, done that.
For "finding anomalies," though, it is far from the only framework available--and the critical bit isn't the building blocks but the structure you build with it.
I used parts of ROOT in building a data acquisition system--I needed to fit for the center of a laser beam. I used different parts of it to create a tool for studying data provided in text format--a fancier version of TOPDRAWER. (All the cool kids are using jupyter these days, but there's a lot of overhead getting it started. I like lightweight and quick.) Some of the analyses were physics, some were sysadmin stuff (which systems are causing grief when), and I used it for looking at baseball statistics too.
More interesting would be a story about who was teaming up with the econ people, and which directions they were going to take--boosted decision tree, maybe?
No comments:
Post a Comment