Machine Learning Under the Hood: The Technical Tips, Tricks, and Pitfalls


It’s the age of machine learning. Companies are seizing upon the power of this technology to combat risk, boost sales, cut costs, block fraud, streamline manufacturing, conquer spam, toughen crime fighting, and win elections.

This course will prepare you to participate in the deployment of machine learning – whether you’ll do so in the role of enterprise leader or quant. In order to serve both types, this course goes further than typical machine learning courses, which cover only the technical foundations and core quantitative techniques. This course includes illuminating software demos of machine learning in action using SAS products. However, the curriculum is vendor-neutral and universally-applicable. The contents and learning objectives apply, regardless of which machine learning software tools you end up choosing to work with.

What You’ll Learn

  • Participate in the application of machine learning, helping select between and evaluate technical approaches
  • Interpret a predictive model for a manager or executive, explaining how it works and how well it predicts
  • Circumvent the most common technical pitfalls of machine learning
  • Screen a predictive model for bias against protected classes – aka AI ethics

What you will learn from this course

Module 1 – The Foundational Underpinnings of Machine Learning

                      ➤ P-hacking: a treacherous pitfall

                      ➤ P-hacking: your predictive insights may be bogus

                      ➤ P-hacking: how to ensure sound discoveries

                      ➤ Avoiding overfitting: the train/test split

                      ➤ Why ice cream is linked to shark attacks

Causation is just a hobby — prediction is your job

The art of induction: why generalizing from data is hard

Learning from mistakes: why negative cases matter

Intro to the hands-on assessment (Excel or Google Sheets

Module 2 – Standard, Go-To Machine Learning Methods

                      ➤ Business rules rock and decision trees rule

                      ➤ Pruning decision trees to avoid overfitting

                      ➤ DEMO – Comparing decision tree models

                      ➤ Drawing the gains curve for a decision tree

                      ➤ Drawing the profit curve for a decision tree

                      ➤ Naïve Bayes

Linear models and perceptron’s

Linear part II: a perceptron in two dimensions

Why probabilities drive better decisions than yes/no outputs

Logistic regression

DEMO – Training a logistic regression model

Module 3 – Advanced Methods, Comparing Methods, & Modeling Software

                         ➤ Neural nets: decision boundaries

                                & a comparison to logistic regression

                         ➤ DEMO – Training a neural network model

                         ➤ Deep learning

                         ➤ Ensemble models and the Netflix Prize

                         ➤ Supercharging prediction: ensembles

                                & the generalization paradox

                         ➤ DEMO – Training an ensemble model

                         ➤ DEMO – Autotuning a machine learning model

                         ➤ Compare and contrast: summary of ML methods

                         ➤ Machine learning software: dos and don’ts for

                               choosing a tool

➤  Machine learning software: dos and don’ts for choosing a tool

➤  Machine learning software: how tools vary and how to choose one

➤  Model deployment: out of the software tool and into the field

Uplift modelling I: optimize for influence and persuade by the numbers

➤  Uplift modelling II: modelling over treatment and control groups

➤  Uplift modelling III: how it works – for banks and for Obama

➤  Uplift modelling IV: improving churn modelling, plus other applications

Module 4 – Pitfalls, Bias, and Conclusions

                           ➤ Visualizing why models are inequitable

                           ➤ Justice can’t be colorblind

                           ➤ Explainable ML, model transparency,

                                 and the right to explanation

Conclusions on ML ethics: establishing standards as a form of social activism

Pitfalls: the seven deadly sins of machine learning

Conclusions and what’s next – continuing your learning