Syllabus - Overview

This course introduces practical machine learning model building techniques with a strong emphasis on bringing these models to a production environment.

Course Description (What we're going to do)

• Work with code
  • Introduce proper ways to work with code
• Prepare data
  • We will cover preparing unstructured data for easier model ingestion.
• Extract features
  • Feature engineering and its importance!!
  • Creating reusable pipeline transformations
• Build machine learning models
• Which features are important?
• Which model should I use?
• How do I know if the model is any good?
• Deploy them to production
  • Wrap these models into an online application so predictions can be made in real-time

Student Learning Outcomes

• Organize data and extract meaningful features for producing machine learning models
• Build and analyze machine learning models in a way they can be deployed to production environments
• Deploy machine learning models to a production environment

My goals for this class

• To teach you what they didn't teach me when I was in school
• All the topics I want to cover I learned on my own or at my various jobs while going to school
• This class is ambitious
• This class is HARD
  • You'll need to work hard
  • Not an easy ride
  • We will move fast
  • Do not fall behind
• This class is similar to a boot camp

Prerequisites

Equipment

• Computer with a Linux / Bash environment required
  • Ubuntu (or any other Linux Distro) 😍
  • OSX comes with this baked in 😃
  • Windows 10 has a Linux subsystem 😨
  • VirtualBox and install a VM 😐

Required Textbooks

• We are going to touch so many topics, I don't know of "a" good book
• Renaissance man / woman
• Web articles
• Books I've read recently
  • Not required

Syllabus - Homework

• We are each going to build a library that we can use to aid in building machine learning models
• Throughout the semester you will be assigned new features or abilities to add to this library
• Assignments may be directly or indirectly related to building out this library
• These assignments will be submitted as pull requests (PRs) in Github
• You will need to select a peer / friend / acquaintance in the class to review your PR and give meaningful and insightful comments 🤓
• As a reviewer, if you think changes need to be made to the code, you can and should reject the PR asking for fixes
  • If you do reject, make sure to re-review once changes have been made
• You will be graded as both a reviewer and a submitter of PRs
• The instructor will do a final review of the PR and a grade will be determined based off of this

Syllabus - Homework (cont.)

• It is important not to fall behind on these tasks as they will build on each other!
• You will rarely get a PR accepted on its first try
• Other eyes will always see something you’ve missed
• PR must be created and submitted for review on the Wednesday before the homework is due.
  • Gives 🤓 time to review
• Make sure code runs!!
  • I will not be forgiving if it doesn't
• Branches must have the correct (case sensitive) name
  • If it doesn't, automated download processes won't find it and you'll get a 0
• PR must be finished by midnight on the due date

Final Project - What to do

• Pick a fun dataset (here's a few) and predict something of importance from this dataset.
• I will provide some
• I must approve dataset
• You'll need to make a good case if you want to use your own
• Most likely: I won't approve it, but sometimes I do
• You will
  • Analyze the dataset
  • Do feature engineering
  • Build predictive models
  • Analyze model performance

Final Project - What you should code

• Provide a way for the me (instructor) to run this code from beginning to end
  • Using: docker-compose
  • I should be able to
    • Check out Github code
    • run a single script that will download the necessary data, go through the workflow of cleaning, organizing, and preparing the data
    • do all the necessary transformations
    • generate whatever results you need
    • build a predictive model
    • output the predictive model
    • ...

Final Project - Written Report

• Discuss what you wanted to predict. What's in the source dataset
• Discuss data conversions: Cleaned / Organized / ETL (Extract Transform and Load) Procedure
• Feature Analysis
  • Which features were predictive - why?
  • Feature Engineering
    • What did you think would be predictive?
    • Was it? Why or why not? Discuss, even if the feature was a failure
    • Charts / Graphs can be very helpful 📊
• Model building
  • Which models did you try? Why? Results?
  • Accuracy between these models?
  • Final model performance metrics

Final Project - Written Report (cont.)

• The final report should be of sufficient length to cover your modeling process in enough depth that someone without access to your code could recreate your work
• A final report must be a Wiki on your github project

Final Project - Presentations

• You will each present your project to the class
• We'll see how many students are still in the class and allocate time accordingly
• I'll randomly assign time slots ON THE DAY(S) OF PRESENTATIONS!
  • Need to be prepared on the first day
  • Volunteers welcome

Spring Syllabus - Weekly Schedule (subject to change)

Week 10

• Spring Break 🏖️ No Class
  • I'll be available for questions and help

Last Week

• Final Project Presentations

• BDA Syllabus
  • Outdated
  • But printable