4 Predictive AI

 

This chapter covers

  • The iterative process of machine learning
  • Unsupervised and supervised learning
  • Time series analysis and trend detection
  • Personalization through recommendations

While AI is often hyped as a “new” technology, all of us have been consuming it for years and on a daily basis—think Google search, your (imperfect) spam filter, or the entertainment recommendations you get and follow on Netflix or YouTube. Often, we forget about the AI that powers these applications because it runs in the background and doesn’t bother us with too many mistakes. Predictive AI is at work in these applications—a class of algorithms that distill valuable insights from large data quantities. For example, they bring structure into unstructured data, classify data points into meaningful categories, and uncover patterns and relationships that are invisible to humans.

Many companies today skip directly to generative AI, overlooking predictive AI as the critical foundation for data-driven decision making and operations. They sit on a wealth of data but fail to activate it for their business, relying on static knowledge, individual past experiences, and subjective gut feeling. By contrast, a data-driven organization uses large-scale data about its operations, stakeholders, and the larger market context, adding confidence and objectivity to its decisions and actions.

4.1 Unsupervised learning

4.1.1 Using clustering for behavioral segmentation

4.1.2 Preparing training data for clustering

4.1.3 Selecting and training a clustering model

4.1.4 Evaluating clustering models

4.1.5 Optimizing the clustering algorithm

4.1.6 Acting on clustering outputs

4.2 Supervised learning

4.2.1 Preparing training data for classification

4.2.2 Selecting and training a classification model

4.2.3 Evaluating and optimizing the classification model

4.2.4 Acting on classification outputs

4.3 Time series and trend analysis

4.3.1 Adding the time dimension to your data

4.3.2 Extracting meaning from time series data

4.3.3 Acting on time series insights

4.4 Personalized recommendations

4.4.1 Types of recommendation algorithms

Summary