How Much Data Is Enough to Build a Machine Learning Model

Explore techniques for determining the appropriate amount of data needed to build effective machine learning models in this 26-minute video by Jeff Heaton. Learn about extrapolation and interpolation in both univariate and multivariate contexts, and understand how to measure data coverage across multiple dimensions. Discover methods for recognizing multimodal distributions, interpreting machine learning curves, and using Mahalanobis distance. Examine a practical example using a diabetes dataset, including feature importance ranking and creating bounding hyper-rectangles. Gain insights into ensuring your training data adequately represents the full range of scenarios your model may encounter in real-world applications.

Intro
HOW MUCH TRAINING DATA DO YOU NEED?
UNDERSTANDING EXTRAPOLATION AND INTERPOLATION
MULTIVARIATE EXTRAPOLATION
EXTRAPOLATION AND INTERPOLATION IN HIGH DIMENSIONS
MODEL DESIGNED FOR EXTRAPOLATION OR INTERPOLATION
RECOGNIZING MULTIMODAL DISTRIBUTIONS
MACHINE LEARNING CURVE
MAHALANOBIS DISTANCE
EXAMPLE DATASET
DIABETES DATASET
FEATURE IMPORTANCE RANKING
DETERMINE HIGH AND LOW VALUES FOR
CREATE A BOUNDING HYPER-RECTANGLE
MOST DISTANT EDGES OF BOUNDING HYPER- RECTANGLE