Postmodeling & Crosstabs Configuration
Postmodeling Configuration#
The Triage Postmodeling module is controlled by two config files: postmodeling_config.yaml
and postmodeling_crosstabs.yaml
.
Postmodeling Configuration File#
Configuration for the Triage Postmodeling module. An example postmodeling_config.yaml
file can be found here.
project_path
: Project path defined in triage with matrices and modelsaudition_output_path
: Audition output pathmodel_group_id
: List of model_id's [optional if a audition_output_path is given]thresholds
: Thresholds for defining positive predictionsbaseline_query
: SQL query for defining a baseline for comparison in plots. It needs a metric and parametermax_depth_error_tree
: For error trees, how depth the decision trees should go?n_features_plots
: Number of features for importancesfigsize
: Default size for plotsfontsize
: Default fontsize for plots
Postmodeling Crosstabs Configuration File#
Configuration for crosstabs in Triage's Postmodeling module. An example postmodeling_crosstabs.yaml
file can be found here.
output
: Define the schema and table for crosstabsthresholds
: Thresholds for defining positive predictionsentity_id_list
: (optional) a list ofentity_ids
to subset on the crosstabs analysismodels_list_query
: SQL query for gettingmodel_id
sas_of_dates_query
: SQL query for gettingas_of_date
smodels_dates_join_query
: don't change the default query unless strictly necessary. It is just validating pairs of (model_id
,as_of_date
) in a predictions tablefeatures_query
: features_query must joinmodels_dates_join_query
with 1 or more features table usingas_of_date
predictions_query
: the predictions query must returnmodel_id
,as_of_date
,entity_id
,score
,label_value
,rank_abs
andrank_pct
. It must joinmodels_dates_join_query
using bothmodel_id
andas_of_date
.