Blog
Topics related to Data Science. Posts 1 through 5 describe projects completed while enrolled in the spring 2015 data science bootcamp at Metis1.
-
9. Forecasting Wind Power
April 30, 2017
Data collected on the power output of an array of wind turbines provide an interesting case study for machine learning algorithms. I tried a random forest from scikit-learn, the gradient boosting algorithm XGBoost, and a generalized linear model from the R package GAMLSS. Choosing the best model can be surprisingly tricky... -
8. Taking the Confusion out of the Confusion Matrix
July 26, 2016
How Bayes' theorem helps organize the bewildering array of performance metrics that can be estimated from a classifier's confusion matrix. -
7. Predicting Flight Delays with a Random Forest
July 6, 2016
Following up on a workshop on random forests organized by the NYC meetup group Women in Machine Learning and Data Science. -
6. The Eternal Sunshine of Causal Thinking
May 9, 2016
A review of Samantha Kleinberg's latest book, "WHY: A Guide to Finding and Using Causes". -
5. Metis Project Kojak: The Practice of Yoga in the City
June 25, 2015
See how yoga businesses compare in New York City and Los Angeles ...or go straight to the maps: -
4. Metis Project Fletcher: Machine Learning with Tweets
May 29, 2015
How to identify topics by clustering documents, or cluster documents by modeling topics... -
3. Metis Project McNulty: Estimating the Risk of Heart Disease
May 15, 2015
What a single data set can teach us about indicators of heart disease... -
2. Metis Project Luther: Movies and Stars
April 24, 2015
Can we use the star value of a movie's cast to predict revenue? -
1. Metis Project Benson: MTA Data
April 10, 2015
MTA data help a green-energy company deploy sign-up teams in the city...
1. Metis was a 12-week accredited data science bootcamp with campuses in New York City, San Francisco, Chicago and Seattle. It operated from 2014 to 2022. Metis project names were taken from famous detectives on TV shows: Theo Kojak, Jessica Fletcher, Jimmy McNulty, John Luther, and Olivia Benson. The projects themselves were not related to the shows, but the idea was, presumably, to suggest commonalities between good data scientists and good detectives...↩