Caltrain Predict

Caltrain Predict

Andrew Lovett-Barron

Consent management platform for researchers

GA Data Science certificate course

This project is an attempt to analyze twitter (and other) datas to understand whether I can detect disruption within the Caltrain system, and map (with some degree of accuracy) the probability that something will go wrong.

All notebooks available here:

  1. 00getdata - Download and transform twitter data
  2. 01sepEvents - Separate tweets into unique events
  3. 03explore - Initial poking around
  4. 03merge_hand_truth - Merge in hand truth data, truth_tweets.csv
  5. 04fill_in_positives - Take all_stops_in_pa.csv and transform into positives data set
  6. 05merge_with_positives - Merge in positives set
  7. 06initial_analysis - Sketchpad for early interprtetive models
  8. 07focus_decision_tree - Complete analysis: Decision trees and gradient boosting, as well as multiple predictive approaches and tuning.

Sign up for the newsletter

Monthly updates from Andrew Lovett-Barron, mostly writing about design practice, theory, and projects. Occasionally, I may link out to a new project.