README.md

GSP709 —— Identifying Bias in Mortgage Data using Cloud AI Platform and the What-if Tool

Table of Contents （🔎 Click to expand／collapse）

• Overview
• Deploy Models to AI Platform
• Using the What-if Tool to Interpret Model
• References

Overview

In this lab, you use the What-If Tool to identify potential biases in a model that was trained on a mortgage loan applications dataset.

Objectives:

• Build a binary classification model using XGBoost.
• Deploy the model to Cloud AI Platform.
• Use the What-If Tool on the deployed model to search for biases.

Deploy Models to AI Platform

# create a model
$ gcloud ai-platform models create <MODEL_NAME> --regions=<REGION>

# create a version
$ gcloud ai-platform versions create <VRESION_NAME> \
    --model=<MODEL_NAME> \
    --framework=<FRAMEWORK> \
    --runtime-version=<RUNTIME_VERSION> \
    --origin=<MODEL_BUCKET> \
    --staging-bucket=<MODEL_BUCKET> \
    --python-version=<PYTHON-VERSION> \
    --project=<PROJECT> \
    --region=<REGION>

References

• What-If Tool
• AI Platform Documentation
• GoogleCloudPlatform/training-data-analyst | GitHub