task:
  id: grid_functional_zone_classification
  type: episode_classification
  description: Predict TaxiBJ grid functional-zone class for each grid episode.

inherit_general: true

sample:
  primary_key: episode_id

inputs:
  tabular_static:
    source: taxibj_grid_static_metadata
    fields:
      - row
      - col
      - lat
      - lon
  time_series:
    - fields: [inflow, outflow, temperature, windspeed, weather_code, is_holiday]
      source: grid_multivariate
      window:
        length: 672
        unit: 30min

  graph:
    source: grid_8_neighbor_graph

prediction:
  field: zone_id

leakage:
  forbidden_input_columns:
    - target_class_id
    - target_class_name
    - zone_id
    - zone_name
    - episode_id
    - grid_id

data_refs:
  tabular_static:
    taxibj_grid_static_metadata:
      file: TaxiBJ_protocol/input_multimodal/meta_features.csv
      id_column: episode_id
  time_series:
    grid_multivariate:
      file:
        - TaxiBJ_protocol/input_multimodal/time_series/inflow/inflow.h5
        - TaxiBJ_protocol/input_multimodal/time_series/outflow/outflow.h5
        - TaxiBJ_protocol/input_multimodal/time_series/temperature/temperature.h5
        - TaxiBJ_protocol/input_multimodal/time_series/windspeed/windspeed.h5
        - TaxiBJ_protocol/input_multimodal/time_series/weather_code/weather_code.h5
        - TaxiBJ_protocol/input_multimodal/time_series/is_holiday/is_holiday.h5
      value_field: [inflow, outflow, temperature, windspeed, weather_code, is_holiday]
      native_grid_minutes: 30
  graph:
    grid_8_neighbor_graph:
      file: TaxiBJ_protocol/input_multimodal/node_level_task_graphs/grid_8_neighbor_graph/edge_infos.csv