Finding Pokemon using KDTrees in Apache Spark
I often use KDTrees at work, to find which businesses are near a specific location. I wrote this notebook to demonstrate how KDTrees work, and give an example of how you can use them to find Pokemon. In the second half, I show how to use KDTrees inside Spark, and do some optimization of the algorithm. I gave talks on this project at the Puget Sound Python Meetup, and CUGOS.
Machine learning in League of Legends
I got game information for the game League of Legends from the game developer's API using python requests. I then parsed the JSON's to extract features like gold difference, or towers destroyed, and made pandas dataframes of over 200,000 games. Finally, I used random forests in scikit-learn to predict the winners of games before they ended. I wrote an initial blog post describing initial findings, as well as a followup about how skill and region influence predictability. I gave a lightning talk to Seattle's python meetup in December.
Mechanisms Twitter bot
Scientists often use the phrase, "the mechanisms by which X occur are not understood." I wrote a Twitter bot in python using tweepy to query PubMed daily, and find abstracts using that syntax. Then it tweets one abstract per day. Runs on an AWS instance. I wrote a short blog post about analytics of the mechanisms phraseology.
As a final project for a data science class, I used multiple types of regression (including bootstrapped and LASSO) algorithms on hubway bikeshare data, using R. Major findings were that people like to bike downhill, weather negatively impacts the number of rides people take, and that casual riders take longer trips than registered members.
This is the python code I use in my day-to-day work. Of interest might be the GUI I made to help other people in the lab study feeding behaviour, BioDAQ GUI.
This repository shows select code from a classifier I made in MATLAB to analyze olfactory neuron behaviour.