Predictive Analytics: Course Notes, Tests, and Assignments

Welcome! This repository contains the teaching material that I used for a few courses on Predictive Analytics that I taught in 2020-2021 (AFM 346, AFM 416, ACC 690 at the University of Waterloo). The repo includes course notes, assignments, take-home examinations, and other material, including suggested responses for assignments and exams. It does not include discussion questions that we used for live sessions, nor recordings of our lectures.

The course lectures were organized as shown in the outline below. The assignments include solutions in files typically labelled with the suffix _code. They also include my guidance notes for TAs reviewing the assignments. I have also included the recommended readings which are available online.

I publish the material as is and, unfortunately, do not have the capacity to provide support to users (answer questions and so on). However, if you find an error or enhancement, feel free to highlight it using Github Issues.

1. Getting Started

Recommended readings:

• Healy: Data Visualization.

  • Look at data
  • Get started

• Khun and Johnson. Feature Engineering and Selection: A Practical Approach for Predictive Models.

  • Exploratory Visualizations

Includes Assignment 1.

Predictive Analytics and Machine Learning

Recommended readings:

• Cimentada. Machine Learning for Social Scientists.

  • Machine Learning for Social Scientists

• Khun and Silge. Tidy Modeling with R.

  • Software for modeling

• Boehmke and Greenwell. Hands-On Machine Learning with R.

  • Modeling Process

Linear Regression, K Nearest Neighbours, and Spending Your Data

• Khun. Applied Machine Learning - Basic Principles.

• Tidymodels.org. Get Started.

  • Build a model

• Khun and Silge. Tidy Modeling with R.

  • Spending our data

• Boehmke and Greenwell. Hands-On Machine Learning with R.

  • Linear Regression
  • Logistic Regression

• Ismay and Kim. Statistical Inference via Data Science.

  • Basic Regression
  • Multiple Regression

Includes Assignment 2.

Feature Engineering and Workflows

• Boehmke and Greenwell. Hands-On Machine Learning with R.

  • Feature & Target Engineering

• Khun and Johnson. Feature Engineering and Selection: A Practical Approach for Predictive Models.

  • Encoding Categorical Predictors
  • Engineering Numeric Predictors
  • Handling Missing Data

• Tidymodels.org. Get Started.

  • Preprocess your data with recipes

• Khun and Silge. Tidy Modeling with R.

  • Feature engineering with recipes
  • Fitting models with parsnip
  • A model workflow

Includes Assignment 3.

Measuring Performance

• Cimentada. Machine Learning for Social Scientists.

  • Loss Functions

• Khun and Silge. Tidy Modeling with R.

  • Judging model effectiveness

Resampling Methods

• Khun and Silge. Tidy Modeling with R.

  • Resampling for evaluating performance
  • Comparing models with resampling
  • Model tuning and the dangers of overfitting

Regularization and Hyperparameter Tuning

• Cimentada. Machine Learning for Social Scientists.

  • Regularization

• Boehmke and Greenwell. Hands-On Machine Learning with R.

  • Regularized Regression

• Khun and Johnson. Feature Engineering and Selection: A Practical Approach for Predictive Models.

  • Feature Selection Overview

Decision-Trees and Ensemble Methods

• Cimentada. Machine Learning for Social Scientists.

  • Tree-based methods

Includes Assignment 4.

Support Vector Machines

• Boehmke and Greenwell. Hands-On Machine Learning with R.

  • Support Vector Machines

Neural Nets

• Boehmke and Greenwell. Hands-On Machine Learning with R.

  • Deep Learning

Includes Assignment 5.

Creating End-to-End ML Experiments

• Khun and Johnson. Feature Engineering and Selection: A Practical Approach for Predictive Models.

  • A Review of the Predictive Modeling Process

• Khun and Silge. Tidy Modeling with R.

  • Appendix A Recommended preprocessing

Copyright and License

Copyright © 2020-2021 Jesús Calderón.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.