Because of academic restrictions and consulting agreements, I am unable to share complete versions of my projects publicly.
These folders contain example code from the following projects:
Utilizing historical equity and options data to predict pending acquisitions utilizing a deep LSTM model with weighted classes, batch normalization, and dropout in Keras.
Utilizing a variety of NLP tools, I process text files from an Amazon S3 bucket to extract hypothesized events that can be verified by Mechanical Turks. These are then used to develop a deep seq2seq model extracting the event name and date (not included here).
In my NLP class for the MIDS degree at UC Berkeley, my team developed a bi-directional LSTM model to identify the type and location of grammatical errors in text.
For my team's capstone project at UC Berkeley, I developed hand-extracted features of tweets that indicated their actionability and used a set of hand-classified tweets to train a bi-directional LSTM.