Visualization Best Practices for Explainable AI

Dive deeper into understanding and explaining machine learning terms and charts to business stakeholders using Jupyter Notebooks, Python, and Power BI in this 58-minute conference talk from PASS Data Community Summit. Explore real-world examples from healthcare, retail, marketing, banking, and other industries while learning about the differences between machine learning and artificial intelligence, automated machine learning processes, and the importance of explainable AI. Discover techniques for visualizing model workings, investigating training data, exploring variable relationships, examining decision rules, analyzing prediction explanations, evaluating model performance, and detecting model changes over time. Gain valuable insights into visualization best practices for explainable AI to effectively communicate complex machine learning concepts to non-technical audiences.

Intro
Technical Assistance
Machine Learning vs. Artificial Intelligence (AI)
Why do they call it "Machine Learning"? Machines learn from data to make predictions on new data
Two stage machine learning process How does it work?
Automated Machine Learning (AutoML) 3 simple steps
Explain results Need to speak data using a common language
Explainable Al, is it possible?
Helps avoid incompleteness issues Explanations fundamentally help identify gaps in problem formalization - incompleteness
Visualize how models work
How to understand a model
What data did the model use? Understand model data source limitations
Investigate model training data
Understand what matters
Explore relationships between variables
Examine decision rules Rules Fit Classifiers
Discover business rules from text fields
Analyze prediction explanations Global
Evaluate model performance
Understand how well a model fits the data
Examine where models make mistakes
Detect model changes over time
Time series
Visualize probability Workflows
Resources