NYC Taxi Trip Duration — Tutorial

A step-by-step guide to reproduce experiments and get a submission.

1. Setup

• Install Python 3.8+ and create a virtualenv: python -m venv .venv source .venv/bin/activate pip install -r requirements.txt

2. Data

• Download Kaggle dataset and place files: data/raw/train.csv data/raw/test.csv

3. Preprocessing

• Run: python src/preprocess.py
  --train data/raw/train.csv
  --test data/raw/test.csv
  --out data/processed/

• What it does:

  • Parses datetimes, fills missing values
  • Adds features: hour, dow, haversine_distance, bearing
  • Saves parquet files in data/processed/

Files to inspect

• notebooks/ — walkthroughs
• src/preprocess.py — transformation pipeline
• src/train.py — training configs and CV logic