Hi :)

Our code is entirely in /home/ec2-user/environment. The main script that we execute is dual_listing.py. It imports stuff from Recommendation.py (which includes classes used to interface with the new processing algorithm) and analyser_news.py, which contains the new analyzing algorithm.

We try to analyze news as well as exploit the dual-listing situation. For the former, we did not succeed in training our own network to give good predictions, so our algorithm consist of a long lookup table that associates adjectives with a "positivity score". The latter is considered in the amount of lots that we buy or sell. If an instrument's dual partner is cheaper, we only sell lots (instead of the usual buy and sell). Conversely, if the dual is more expensive, we only buy lots of the instrument, but sell none.