Quick Start¶

Basic Usage¶

Run the prediction pipeline with the default model:

python run.py

Selecting a Model¶

Choose a specific model:

# Markov chain model
python run.py --model markov

# Temporal flow model
python run.py --model temporal_flow

# Persistence baseline
python run.py --model persistence

# Station average model
python run.py --model station_average

Setting Random Seed¶

For reproducible results:

python run.py --model markov --seed 42

Configuration¶

Edit config.yaml to customize:

Data paths and time ranges
Cross-validation parameters
Model-specific settings
Empty/full station thresholds

Example configuration:

data:
  trips_path: "data/parquet/trips"
  station_info_path: "data/parquet/stations/station_info.parquet"

model:
  name: "markov"
  markov:
    smoothing_alpha: 0.0
    min_transitions: 10
    n_simulations: 1
    random_seed: 42

cross_validation:
  n_folds: 4
  train_window_days: 7
  test_window_hours: 24