Neighboring Values (Up and Down)

The flank_up and flank_down features are observation-level features that represent the neighboring values of an observation, such as temperature and salinity. Although any columns in the input dataset can be specified for these features, they are usually coupled with the variables used in the basic_values feature.

Configuration: Setup

To include the flank_up and/or flank_down features in your training and classification datasets, the values flank_up and/or flank_down need to be specified in the feature_sets section.

feature_sets:
  - name: feature_set_1
    features:
      - flank_up
      - flank_down

Configuration: Parameters

Both flank_up and flank_down features require two common mandatory parameters (col_names and stats_set) and one feature-specific parameter (flank_up or flank_down).

• The col_names parameter specifies the column names in the input dataset that will be used for the flank_up and flank_down features.

• The stats_set parameter specifies how the feature values are normalized. aiqclib currently supports raw and min_max as normalization methods. The name value in stats_set must correspond to a name in the feature_stats_sets section.

• The flank_up and flank_down parameters specify the number of neighboring values to include in the feature.

feature_param_sets:
  - name: feature_set_1_param_set_1
    params:
      - feature: flank_up
        col_names: [temp, psal, pres]
        stats_set: { type: min_max, name: basic_values3 }
        flank_up: 5
      - feature: flank_down
        col_names: [temp, psal, pres]
        stats_set: { type: min_max, name: basic_values3 }
        flank_down: 5

Configuration: Normalization

If the normalization method is not set to raw, the summary statistics specified here will be used for normalization. These features normally use the same summary statistics as the corresponding basic_values feature.

feature_stats_sets:
  - name: feature_set_1_stats_set_1
    min_max:
      - name: basic_values3
        stats: { temp: { min: 0, max: 20 },
                 psal: { min: 0, max: 20 },
                 pres: { min: 0, max: 200 } }

Note

aiqclib offers helper functions to calculate summary statistics (like min/max values). Please refer to the Feature Normalization guide for details.