This module is used to collect the distance-based metrics used to evaluate the novelty content of synthetically generated datasets.

The module evaluates the Gower’s distance of each row of x_dataframe with respect to each row in y_dataframe, eventually leading to information such as Distance to Closest Record (DCR) and DCR Share (see validation_dcr_test() below for more information).

APIs

distance_to_closest_record()

dcr_stats()

number_of_dcr_equal_to_zero()

validation_dcr_test()

Examples

The argument dcr_name, **given as the first argument in most of these functions, serves as an indication for the report produced with the function report().

The functions accept one of the following for the dcr_name argument:

• "synth_train": indicating that the function is computing or manipulating the distance to closest record between the synthetic dataset and the real one used for training the generative model with which the synthetic dataset was produced.
• "synth_val": indicating that the function is computing or manipulating the distance to closest record between the synthetic dataset and the real holdout one used for validation (NOT employed for training the generative model with which the synthetic dataset was produced)
• “other”: in this case the result of the function will not be stored in the json file used to generate the report.

It is possible to specify the path where to save the new information computed in a json file.

dcr_synth_train       = distance_to_closest_record("synth_train", 
																									 synth_data, 
																									 real_data, 
																									 path_to_json="path/to/json/")
																									 
dcr_synth_valid       = distance_to_closest_record("synth_val", 
																									 synth_data, 
																									 valid_data, 
																									 path_to_json="path/to/json/")
																									 
dcr_stats_synth_train = dcr_stats("synth_train", 
																	dcr_synth_train, 
																	path_to_json="path/to/json/")
																	
dcr_stats_synth_valid = dcr_stats("synth_val", 
																	dcr_synth_valid, 
																	path_to_json="path/to/json/")
																	
dcr_zero_synth_train  = number_of_dcr_equal_to_zero("synth_train", 
																										dcr_synth_train, 
																										path_to_json="path/to/json/")
																										
dcr_zero_synth_valid  = number_of_dcr_equal_to_zero("synth_val", 
																										dcr_synth_valid, 
																										path_to_json="path/to/json/")
																										
share                 = validation_dcr_test(dcr_synth_train, 
																						dcr_synth_valid, 
																						path_to_json="path/to/json/")

Note that if the json file already exists in the specified directory, the new information is appended or, if already present in the file, updated.