This is my capstone project for the Data Incubator Fellowship. My goal is to take this project from initial concept to a usable application.
- I have found and scraped several investor news sites for past and upcoming FDA PDUFA (Prescription drug user fee act) action dates
- I have munged the FDA announcement dates and joined them to stock price time-series for the relevant companies
- I have sliced the relevant data into 120-day long preceding and following periods for data annotation and model training
- I have annotated the data
- I have calculated and selected important features from my training data
- I have selected and trained a classification model (Support Vector Classifier)
- I have cross validated the model, and verified it preforms better than random or majority case guessing
- I have made a basic visualizaton showing how the model can trade better than a lay person
Check out the approximate portfolio values for my model's trading recomendations:
SVM Advised Trades in Blue
Random Trades in Green
All Trades in Red
- Make a server (new repo) to constantly monitor the market and make trade decisions
- Learn webapp development the hard way
- Make a real project with a real repository on GitHub so I have something to demonstrate my skills beyond the lackluster scripts I wrote in graduate school