Classifying train and car journeys using telematics data
summary
A working classifier distinguishes train versus car trips from MyDrive Solutions telematics by engineering features like percentage of points snapped to roads, path efficiency and speed variability, extracted from second‑by‑second GPS stored on S3 and processed with Spark SQL/Zeppelin and R (RStudio, geosphere, leaflet, caret). A logistic regression deployed in the scoring pipeline achieves strong separation (reported ≈86% AUC), random forests improve accuracy, and the system successfully drops train trips from driver scoring while needing more labelled data across countries.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 7:43Need for train-car classifier
- 13:37Telematics data features
- 21:44Modeling and results
- 28:58Implementation lessons
- 35:15Q&A
related talks
40mSmart cars of tomorrow: real-time driving patterns
27mUsing Machine Learning to solve a classification problem with scikit-learn
19mDrive transformation through machine learning
45mFinding Driving Style Patterns in Caterpillar Machine Data
44mA Practical-ish Introduction to Data Science
42mCreating correct and capable classifiers
30mCrash detector
23mDataEngBytes 2023 - MEL-T3-01 - Rajath Vanikul