config.yaml

model:
  name: rossmann_sales_predictor
  description: Predict daily Rossmann store sales with a small XGBoost forecasting pipeline.
  tags: ["time-series", "regression", "sales-prediction", "xgboost"]

data:
  features:
    - "Store"
    - "DayOfWeek"
    - "Promo"
    - "StateHoliday"
    - "SchoolHoliday"
    - "Year"
    - "Month"
    - "Day"
    - "IsWeekend"
    - "StoreType"
    - "Assortment"
    - "CompetitionDistance"
    - "LogCompetitionDistance"
    - "Promo2"
    - "Promo2SinceWeek"
    - "Promo2SinceYear"
    - "days_to_easter"
    - "easter_effect"
    - "fourier_sin_1"
    - "fourier_cos_1"
    - "fourier_sin_2"
    - "fourier_cos_2"
    - "fourier_sin_3"
    - "fourier_cos_3"
    - "fourier_sin_4"
    - "fourier_cos_4"
    - "fourier_sin_5"
    - "fourier_cos_5"
  target: "Sales"
  train_path: "./data/raw/train.csv"
  store_path: "./data/raw/store.csv"

pipeline:
  fourier_period: 365.25
  fourier_order: 5

model_params:
  xgboost:
    n_estimators: 500
    learning_rate: 0.05
    max_depth: 10
    subsample: 0.8
    colsample_bytree: 0.8
    random_state: 42
    n_jobs: -1
    verbosity: 0
    objective: "reg:squarederror"