Inspiration
Merging or acquisitions
What it does
We generate two signals:
- one to predict the likelihood of a merging, small acquisition or big acquisition event.
- one forecasting the stock price evolution under a pre/post-acquisition regime.
How we built it
The first one uses the CEO's twitter feeds coupled with the earning calls transcripts and fundamental financial data related to the underlying price stock.
Challenges we ran into
First, we were lacking almost all the data; we had to scrap it from several different places and complete it using other datasets.
Second, we initially worked with a BERT word embedding and wanted to train but its dimensionality being to high for the number of points we were considering, we had to consider a theoretically simpler approach (sentiment analysis).
Accomplishments that we're proud of
Yes.
What we learned
Limitation of the usage of BERT embedding in the training for small datasets.
What's next for Predicting acquisitions
Making text feed richer, optimising the way the signal are coupled together to generate the final buy/sell signal.
Log in or sign up for Devpost to join the conversation.