I built a Deep Learning Network for NBA game prediction. Game data from 2004-2020 were used. Please see the GitHub Repo for the files as well as for a PowerPoint presentation describing the various optimization techniques I experimented with to improve network accuracy. Generalization accuracy (on a test set) peaked at around 62% which is below Vegas betting odds of around 68%.

GitHub Repo found here: https://github.com/hermessuen/DeepNetsForNBA

The purpose of the project is to correctly classify heartbeats from ECG data as either "abnormal" or "normal". Specifically, "abnormal" heartbeats will refer to heartbeats with Arrythmia. From there, we want to apply several Machine Learning Interpretability algorithms to our models to see what features they are actually using to make their predictions.

Th project is split up into two parts: a MATLAB live script and a MATLAB AppDesigner application.

GitHub Repo found here: https://github.com/hermessuen/ArrythmiaDetection

This project was inspired by the wave of protests supporting the Black Lives Matter movement that occurred in the spring and summer of 2020. In an effort to stop the spread of misinformation and to use twitter for good, I applied for a Twitter development account. I used this account to stream tweets from Twitter containing “blacklivesmatter”, “blm” in conjunction with words like “protest”, “sign”, and “donate”. The goal was to build a machine learning classifier that could tweet out periodically useful information regarding social justice, such as times and locations of protests, places to donate, and petitions to sign. Features were extracted using the TFID vectorizer from sklearn (which finds weightings of words based on the number of documents they appear in so as to capture highly important words). An SVM classifier was then trained on a dataset with 20% hold out for testing.

The classifier achieved about 97% accuracy and now the account tweets out every hour with useful information on places to donate, locations to protest, and petitions to sign.

GitHub repo found here: https://github.com/hermessuen/tweetForActivism