| # ============================================================ | |
| # Task: lane_count_regression | |
| # ============================================================ | |
| # Task objective: | |
| # - Predict the lane count, Lanes, from static attributes of LargeST road segments / sensors. | |
| # - This is an episode-level regression task: each ID corresponds to one static sample. | |
| # | |
| # Core semantics: | |
| # - sample is a road_segment / sensor ID. | |
| # - inputs are the static attributes available to the model. | |
| # - prediction.target.column = Lanes is the regression target the model should predict. | |
| # - Lanes must not be an input field, as this would directly leak the label. | |
| # ============================================================ | |
| task: | |
| id: lane_count_regression # Unique task ID; keep it consistent with the directory name. | |
| type: episode_regression # Task type: output one continuous value or numeric label per sample. | |
| description: > | |
| Predict the number of lanes for each LargeST road segment / sensor episode | |
| using static metadata features. | |
| inherit_general: true | |
| sample: | |
| primary_key: ID # ID joins labels.csv, splits.csv, and the static attribute table. | |
| inputs: | |
| # ------------------------------------------------------------ | |
| # Static tabular inputs. | |
| # ------------------------------------------------------------ | |
| tabular_static: | |
| source: largeST_static_metadata # Refers to the static attribute table in data_refs.tabular_static. | |
| # Explicitly list the static fields available to the model. | |
| # The following fields are common in LargeST / PeMS road metadata and may help predict lane count. | |
| # If the column names in your metadata file differ, update them here accordingly. | |
| fields: | |
| - Lat | |
| - Lng | |
| - District | |
| - County | |
| - Fwy | |
| - Lanes | |
| - Type | |
| - Direction | |
| time_series: | |
| - fields: traffic_flow # Historical time series fields available to the model. | |
| source: traffic_flow_ca_2019 # Refers to the underlying LargeST h5 file defined in data_refs.time_series. | |
| window: | |
| length: 1344 | |
| unit: 15min | |
| graph: | |
| source: road_distance_graph | |
| prediction: | |
| field: Lanes # Supervision column in labels.csv. | |
| leakage: | |
| forbidden_input_columns: | |
| - Lanes # Target column; it must never be used as an input. | |
| - ID # The primary key is used only for joins / indexing. | |
| # If Fwy reveals too much about lane count, move it here from inputs.tabular_static.fields. | |
| # - Fwy | |
| data_refs: | |
| tabular_static: | |
| largeST_static_metadata: | |
| file: All_Datasets/LargeST_ca_full/input_multimodal/ca_meta.csv # Adjust this to the actual file location; relative paths are preferred. | |
| id_column: ID | |
| time_series: | |
| traffic_flow_ca_2019: | |
| file: All_Datasets/LargeST_ca_full/input_multimodal/time_series/traffic_flow/ca_his_2019.h5 | |
| value_field: traffic_flow | |
| native_grid_minutes: 15 | |
| graph: | |
| road_distance_graph: | |
| file: All_Datasets/LargeST_ca_full/input_multimodal/node_level_task_graphs/road_distance_graph/edge_infos.csv |