This is the repository for Practical Machine Learning with R (CSX460) at the University of California, Berkeley. The most recent class is/was Fall 2016.
This course provides an introduction to machine learning using R, the open source, statistical programming language. Once a niche set of tools for statisticians, programmers and quants, machine learning (sometimes also called sttistical learning or data mining) has spread in popularity to become an indispensable tool to a wide variety of applications and disciplines. This course teaches the fundamentals of machine learning without delving into too much mathemtics or code. The course will teach practical aspects of machine learning. Upon completion of the course students will be able to apply lessons to solve problems using machine learning in their own work.
Students of this class will learn:
- Fundamental concepts in ML
- The differnece between supervised, unsupervised, semi-supervised, adaptive/reinforcement learning
- The three prerequisites of ML algorithms/models
- Loss function
- Restricted class of functions
- Search methodology for training
- How to evaluate and compare ML model performance
- How to pre-process data and build features
- How to train ML models for prediction, categorization and recommendations
- How to apply ML models on new data
- How to use resampling techniques to calculate model performance
- What the bootstrap is and how it works
- What Bagging is and how and why it improves model performance
- What Boosting is and how and why it improves model performance
- How to implement/deploy ML models for use by a wider audience
- How to frame questions to be answered using ML techniques
- Collaborate in a group using tools for collaborative/social programming
- Generate high quality, graphical and textual results
- Anyone who wishes to learn the fundamentals of machine learning
- Anyone who wants to learn about using R to build, evaluate or deploy machine learning models.
- Scientists, engineers, business analysts, research who explore and analyze data and wish to present their findings in well-formatted textual and graphical forms.
- Anyone wishing to get hands-on experience building machine learning models.
- Experience programming in at least one high-level programming language such as BASIC, PASCAL, C, Java, Python, Perl, or Ruby.
- Familiarity with R such as that gained through the Programming with R course.
- Basic knowledge of statistics as covered in a first-semester undergraduate statistics course. There will be some coverage of basic statistical techniques as part of covering core elements of the Machine Learning.
- Personal laptop for class assignments.
Text for the Course:
***Machine Learning with R, 2nd Edition***
ISBN: 978-1-78439-390-8
Lantz, Brett
Packt Publishing
2015
Recommended:
***Applied Predictive Modeling***
ISBN-13: 978-1461468486
ISBN-10: 1461468485
Kuhn, Max and Johnson, Kjell
Springer Science+Business
2013
All assignments are due the day before the next lecture.
There is an google group for this class: CSX460. Contact the professor for entrance to the group.
Current Term: Fall 2016
This provides a session by session overview of CS-X460 (Practical Machine Learning).
Topics:
- Welcome and Introductions
- Class Book, Materials, etc.
- Course Overview
- Setting up your environment
- Installing R/R Studio
- Installing git and using Github
- Installing packages from CRAN and Github
- Introduction to Maching Learning
Reading:
- Machine Learning with R (MLR), Chapters 1-2
- Data processing
- Building First Models
- Supervised, unsupervised, and semi-supervised
- Regression and classification
- Measuring model error(s)
- Machine learning prerequisites
- Algorithm types
Reading:
- MLR Chapter 3, Chapter 6 pp.171-200
- Introduction to dplyr
- Introduction to data.table
- Introducting magrittr
Exercise(s): 02-fundamentals/02-exercise-nycflights.Rmd
R Packages Introduced:
- General awesomeness: magrittr
- Data Manipulation: dplyr, data.table
- Reading data: readr, data.table::fread
Topics:
- Linear Regression
- Model Formula
- Modeling Process
Reading:
- Introduction to R Graphics with ggplot2
- ?formula
- ?MASS::stepAIC
- Suggested:
Exercise(s):
- 03-linear-regression/
R Packages Introduced:
Reading:
- Introduction to Statistical Learning, Classification 4.1-4.3 "Logistic Regression"
- MLwR Chapter 10
- Suggested:
Exercise(s):
- 03-exercise-nyc-flights-logistic.Rmd
R Packages Introduced:
- stats::glm
- MASS::stepAIC
Reading:
- MLwR Chapter 5
Exercise(s):
- classification-metrics-exercises.Rmd
R Packages Introduced:
- caret
- gmodels
- ROCR, pROC
- Decision Trees/Recursive Partitioning
- Bias-Variance Trade-off
- Introduction to Caret
Reading:
- MLwR Chapter 11 Improving Model Performance (First Part)
- Tuning Stock Models (pp. 347-358)
- Review Caret Website There is a lot in the caret website, it is most important to familiarize yourself with the use of the models.
Exercise(s):
- None.
R Packges introduced:
- caret
- rpart
- ctree
- C50
- Bagging
- Bagged Trees / Random Forests
- Boosting
Exercises:
- exercise-caret-models.Rmd (Due: 2016-12-06)
Reading:
- MLwR Chapter 11 Improving Model Performance (Second Part)
- Tuning Stock Models (pp. 359-376)
R Packages introduced:
- caret
Reading:
- CRAN Task View: Time Series Analysis
- Forecasting Principals and Practice
- Chapters 1 "Getting Started"
- Chapter 2 "The Forecaster's Toolbox"
- Chapter 3 "Judgemental Forecasts"
- Chapter 8 "Arima Models"
Exercise(s):
- Complete control-lift in-class exercise
- Complete revenue forecast exercise
R Packages Introduced:
- survival : survReg
- forecast : Arima
Topics:
- Delivery and Production
- Patterns
Reading:
Exercise(s):
- None