AI Verify Examples

Mock Samples

We have included mock samples in this repository. To use these examples, follow our user guide to upload them to the respective pages:

1. Upload all mock datasets in ./data to Datasets via AI Verify user interface.
2. Upload all mock models and pipelines in ./models and ./pipeline to AI Models via AI Verify user interface.

Note that all datasets (except for Fashion-MNIST dataset) are synthetically created using numpy.

Binary Classification

Use Case: To predict whether an applicant will default the loan

Output Classes:

• 0: Will not default the loan
• 1: Will default the loan

Non-pipeline

	File
AI Model	sample_bc_credit_sklearn_linear.LogisticRegression.sav
Test Dataset	sample_bc_credit_data.sav
Ground Truth File	sample_bc_credit_data.sav
Ground Truth	default
Sensitive Feature	gender, race

Pipeline

	File
AI Model	binary_classification_tabular_credit_loan
Test Dataset	sample_bc_pipeline_credit_data.sav
Ground Truth File	sample_bc_pipeline_credit_ytest_data.sav
Ground Truth	default
Sensitive Feature	gender, race

Multiclass Classification

Use Case: To categorise whether a datapoint into its respective toxic category

Output Classes:

• 0: Not toxic
• 1: Racism
• 2: Violence
• 3: Hatred
• 4: Identity Hate

Non-pipeline

	File
AI Model	sample_mc_toxic_sklearn_linear.LogisticRegression.sav
Test Dataset	sample_mc_toxic_data.sav
Ground Truth File	sample_mc_toxic_data.sav
Ground Truth	toxic
Sensitive Feature	gender, race

Pipeline

	File
AI Model	multiclass_classification_tabular_toxic_classification
Test Dataset	sample_mc_pipeline_toxic_data.sav
Ground Truth File	sample_mc_pipeline_toxic_ytest_data.sav
Ground Truth	toxic
Sensitive Feature	gender, race

Regression

Use Case: To predict how much a donor will donate

Output:

• Donation amount

Non-pipeline

	File
AI Model	sample_reg_donation_sklearn_linear.LogisticRegression.sav
Test Dataset	sample_reg_donation_data.sav
Ground Truth File	sample_reg_donation_data.sav
Ground Truth	donation
Sensitive Feature	gender, race

Pipeline

	File
AI Model	regression_tabular_donation
Test Dataset	sample_reg_pipeline_data.sav
Ground Truth File	sample_reg_pipeline_ytest_data.sav
Ground Truth	donation
Sensitive Feature	gender, race

Image - Multiclass Classification

Use Case: To classify images into their respective fashion object (e.g. shoes, clothes)

	File
AI Model	multiclass_classification_image_mnist_fashion
Test Dataset	raw_fashion_image_10
Ground Truth File	pickle_pandas_fashion_mnist_annotated_labels_10.sav
Ground Truth	label
Sensitive Feature	None
Annotated File	pickle_pandas_fashion_mnist_annotated_labels_10.sav
Annotated Columnn	file_name

License: The copyright for Fashion-MNIST is held by Zalando SE. Fashion-MNIST is licensed under the MIT license.