I often find myself getting carried away with data and models with no real goal of making a larger project out of the matter. This is where those types of projects will live. So far I have:
An NLP analysis of the career works of various authors. Here I look into the distribution of part-of-speech and punctuation profiles across authors, cluster them on both linguistic profile and word choice (via TF-IDF scores), and hierarchically cluster all the authors in my data. Data mostly comes from Project Gutenberg.
Every cup of coffee I consumed between July 13th and December 21st, 2016 was logged as an entry in a spreadsheet I keep. This is the analysis of that data.
Some minimal analysis on polling and electoral college data. Still need to add Trump vs. Hillary to this.
Procedural text generation with recurrent neural networks using long short-term memory cells. Makes use of the Keras library.
A notebook for data mining your Groupme. Requires you to get an API key. Find out who talks the most, writes the longest messages, gets the most likes, swears the most, etc. I highly recommend doing this.
Comparing the performance of three binary classifiers (logistic regression, random forest, and SVM) on the UCSD Malicious URL dataset (~1M rows, 3M features).
An analysis of the career value of NBA draft picks, draft-day trades, and of tanking on the league standings in order to improve a team's lottery odds.