Unit 3: Required

This lesson uses the Iris dataset and the NBA player statistics dataset. The Iris dataset allows students to easily make their own rules-based model and is easy to visualize for KNN. The NBA dataset results in a very nice curve for choosing K.

After this lesson, students should be able to:

• Utilize the KNN model on the iris data set.
• Implement scikit-learn's KNN model.
• Assess the fit of a KNN Model using scikit-learn.

Before this lesson(s), students should already be able to:

• Load, explore, and manipulate data using Pandas
• Create simple visualizations with Matplotlib
• Interpret statistical information from box and scatter plots
• Describe the statistical meaning of an "error"

TOTAL (170 min)

• Learning Objectives (5 min)
• Overview of the Iris Data Set (10 min)
  • Terminology
• Exercise: "Human Learning" With Iris Data (60 min)
• Human Learning on the Iris Data Set (10 min)
• K-Nearest Neighbors (KNN) Classification (30 min)
  • Using the Train/Test Split Procedure (K=1)
• Tuning a KNN Model (30 min)
  • What Happens If We View the Accuracy of our Training Data?
  • Training Error Versus Testing Error
• Standardizing Features (15 min)
  • Use StandardScaler to Standardize our Data.
• Comparing KNN With Other Models (10 min)

For more information on this topic, check out the following resources:

• Data School: Machine Learning With KNN
• KNN: Dangerously Simple
• KNN From Scratch
• Detailed Intro to KNN
• Stanford's Machine Learning Course: KNN