7 Supervised learning

 

The (fictional) information technology company JCN Corporation is reinventing itself and changing its focus to artificial intelligence and cloud computing. As part of managing its talent during this enterprise transformation, it is conducting a machine learning project to estimate the expertise of its employees from a variety of data sources such as self-assessments of skills, work artifacts (patents, publications, software documentation, service claims, sales opportunities, etc.), internal non-private social media posts, and tabular data records including the employee’s length of service, reporting chain, and pay grade. A random subset of the employees has been explicitly evaluated on a binary yes/no scale for various AI and cloud skills, which constitute the labeled training data for machine learning. JCN’s data science team has been given the mission to predict the expertise evaluation for all the other employees in the company. For simplicity, let’s focus on only one of the expertise areas: serverless architecture.

7.1    Domains of competence

7.2   Two ways to approach supervised learning

7.3    Plug-in approach

7.3.1   Discriminant analysis

7.3.2   Nonparametric density estimation

7.4   Risk minimization basics

7.4.1   Empirical risk minimization

7.4.2   Structural risk minimization

7.5    Risk minimization algorithms

7.5.1   Decision trees and forests

7.5.2   Margin-based methods

7.5.3   Neural networks

7.5.4   Conclusion

7.6    Summary