Overview

In this course, the learner is guided through a realistic scenario of governing the predictive models of a data science project during their lifecycle. The project focuses on creating machine learning models that can be used for mortgage loan approvals, where decisions highly influence both individuals and organizations. The narrative is driven by Anna Parker Sr. Product manager of a financial institution in charge of mortgage approval and Sr Data Scientist, Sara Man that guides the learner who assumes the persona of a Jr. Data Scientist, Leo Meep through the usage of IBM watsonx.governance to detect bias and monitor their deployed machine learning models for drift in their selected metrics.

The course educates the learner in IBM’s fundamental pillars of trustworthy AI such as explainability, fairness, and transparency and guides the learner through hands-on exercises using the graphical user interface, in creating machine learning models, tracking the model lineage, enriching the model with metadata previously known as AI Factsheets, and exploring the fundamental pillars of trustworthy AI using watsonx.governance. The learner deploys and monitors the model for drift using the graphical user interface of Watson OpenScale.

Audience

Data Analysts, Data Scientists, Business Analysts, and Researchers

Prerequisites

The learner prerequisite skills and knowledge include:

• Experience working in a browser
• Working knowledge of electronic mail including basic mail, calendar, and address book tasks
• Some experience using word processing, presentation, and spreadsheet programs
• Experience working in browser.
• Basic knowledge of machine learning and data science.
• Familiarity with IBM watsonx products would be helpful
• Basic knowledge of the data science process
• Basic knowledge of Jupyter Notebooks, APIs, SDKs and Python.

Objective

After completing this course, the learner will be able to:

• Define the terms governance, bias, fairness, risk, lineage, metadata
• Explain the importance of AI governance
• Distinguish between data governance vs AI governance
• Create an AI use case and associate with an AI model
• Build a deployment space and deploy a predictive AI model
• Evaluate an AI model for drift, bias and fairness using the Insights dashboard
• Choose and configure metrics for an AI model and introduce evaluation data
• Examine model transactions for fairness and explainability

Course Outline

• Introduction
• Module 1: Create a predictive model
• Module 2: Deploy a predictive model
• Module 3: Evaluate a predictive model
• Epilogue