20 Network-driven supervised machine learning

 

This section covers

  • Using classifiers in supervised machine learning
  • Making simple predictions based on similarity
  • Metrics for evaluating the quality of predictions
  • Common supervised learning methods in scikit-learn

People can learn from real-world observations. In some respects, machines can do the same. Teaching computers to metaphorically understand the world through curated experience is referred to as supervised machine learning. In recent years, supervised machine learning has been all over the news: computers have been taught to predict stock prices, diagnose diseases, and even drive cars. These advancements have been rightly heralded as cutting-edge innovations. Yet, in some ways, the algorithms behind these innovations are not that novel. Variants of existing machine learning techniques have been around for many decades; but due to limited computing power, these techniques could not be adequately applied. Only now has our computing strength caught up. Hence, ideas planted many years ago are finally bearing the fruits of significant technological advancements.

20.1 The basics of supervised machine learning

20.2 Measuring predicted label accuracy

20.2.1 Scikit-learn’s prediction measurement functions

20.3 Optimizing KNN performance

20.4 Running a grid search using scikit-learn

20.5 Limitations of the KNN algorithm

Summary