task:
  id: item_category_classification
  type: episode_classification
  description: Predict the item category of each item-store episode using static attributes, historical sales patterns, and item-store graph structure.

inherit_general: true

sample:
  primary_key: episode_id

inputs:
  tabular_static:
    source: meta
    fields:
      - store_id
      - state_id

  time_series:
  - fields: unit_sales
    source: unit_sales_daily
    window:
      length: 56
      unit: day


  - fields: sell_price
    source: sell_price_daily
    window:
      length: 56
      unit: day

  graph:
    source: item_graph

prediction:
  field: cat_id

leakage:
  forbidden_input_columns:
    - cat_id
    - episode_id
    - item_id
    - dept_id


data_refs:
  time_series:
    unit_sales_daily:
      file: m5/input_multimodal/time_series/unit_sales/unit_sales.h5
      value_field: unit_sales
    sell_price_daily:
      file: m5/input_multimodal/time_series/sell_price/sell_price.h5
      value_field: sell_price
    calendar_wday_daily:
      file: m5/input_multimodal/time_series/calendar_wday/calendar_wday.h5
      value_field: calendar_wday
    calendar_month_daily:
      file: m5/input_multimodal/time_series/calendar_month/calendar_month.h5
      value_field: calendar_month
    calendar_year_daily:
      file: m5/input_multimodal/time_series/calendar_year/calendar_year.h5
      value_field: calendar_year
    snap_CA_daily:
      file: m5/input_multimodal/time_series/snap_CA/snap_CA.h5
      value_field: snap_CA
    snap_TX_daily:
      file: m5/input_multimodal/time_series/snap_TX/snap_TX.h5
      value_field: snap_TX
    snap_WI_daily:
      file: m5/input_multimodal/time_series/snap_WI/snap_WI.h5
      value_field: snap_WI

  tabular_static:
    meta:
      file: m5/input_multimodal/meta.csv
      id_column: episode_id

  graph:
    item_graph:
      file: m5-forecasting-accuracy/m5_protocol/input_multimodal/node_level_task_graphs/item_graph/edge_info.csv