Machine Learning with R

# Machine Learning with R
### Bradley C. Boehmke and Brandon M. Greenwell
### 2018/05/12

---

background-image: url(https://i.pinimg.com/originals/e5/26/c1/e526c1b66d356fe88886c1803be31255.png)

# Introductions

---

## Who Are These Guys?

---

## Who Are These Guys?

* 84.51°

* UC, WFU, AFIT

* Dayton, OH

* Married with 2 little girls

* [bradleyboehmke.github.io](http://bradleyboehmke.github.io/)

]

.large[**Brandon Greenwell:**] <img src = "Images/brandon_bitmoji.png" style="width:20%;height:20%;">

* Statistician turned data scientist

* Also at 84.51° (just moved to Cinci from Dayton)

* WSU, UC

* [My R packages](https://github.com/bgreenwell)
 
<img src="Images/food.jpg" width="40%" style="display: block; margin: auto;" />

]

---

background-image: url(Images/intentions.jpg)
background-position: center
background-size: contain

# Intent of the Course

---

## Agenda

* ML Fundamentals

* Regularization

* Multivariate Adaptive Regression Splines (MARS)

* Feed-forward Neural Networks

* Kaggle!

]

* Naive Bayes Classifier

* Support Vector Machines (SVMs)

* Random Forests 🌲

* Gradient Boosting Machines 🌲

* Kaggle!

]

<center>
<bold>

Objective: expose you to a wide-range of techniques and packages that can be used for supervised regression and classification problems.

</bold>
</center>

---

background-image: url(Images/prerequisites.png)
background-position: center
background-size: contain

---

## Basic Assumptions

* Intermediate familiarity with R is required ✅

* This is not your first analytics course ✅

* R installed ( `\(\geq\)` v3.4) ✅

* RStudio installed ( `\(\geq\)` v1.1.4) ✅

* You have a Kaggle acount ✅

<center>
<bold>

Does everyone have the right versions and necessary packages installed?

</bold>
</center>

---

## Packages Used

```r
list.of.packages <- c(
 'alr3', # for Swiss banknote data
 'AmesHousing', # for Ames housing data
 'caret', # for classification and regression training
 'devtools', # for install_github() function
 'dplyr', # for (mostly) pain free data wrangling
 'earth', # for multivariate adaptive regression splines
 'forecast', # for normalizing 
 'gbm', # for generalized boosted models
 'ggplot2', # for awesome plotting
 'ggridges', # for ridgeline plots
 'glmnet', # for elastic net (e.g., the lasso and ridge regression)
 'investr', # for plotFit() function
 'keras', # for deep learning
 'kernlab', # for fitting support vector machines
 'klaR', # for naive Bayes
 'magrittr', # for using the forward pipe operator (i.e., %>%)
 'MASS', # for LDA/QDA functions
 'mda', # for multivariate adaptive regression splines
 'pdp', # for partial dependence plots and ICE/c-ICE curvess
 'plotly', # interactive plots
 'randomForest', # for random forest 
 'ranger', # for fast and efficient random forest
 'RColorBrewer', # for better color palettes
 'rpart', # for binary recursive partitioning (i.e., decision trees)
 'rpart.plot', # for plotting decision tree diagrams
 'rsample', # for data splitting and some data sets
 'svmpath', # for fitting the entire SVM regularization path 
 'tidyr', # for easily tidying data
 'xgboost', # for eXtreme Gradient Boosting
 'yardstick' # for various ML metrics
)

# Run the following lines of code to install the packages you do not have
new.packages <- list.of.packages[
 !(list.of.packages %in% installed.packages()[, "Package"])
]

if(length(new.packages)) {
  install.packages(new.packages)
}

# Install the vip package (for variable importance plots) from our GitHub repo
devtools::install_github("koalaverse/vip")
```
]

---

## Data Used

Although we will demonstrate with several data sets*, our emphasis will be on the following:

```r
# smaller housing data (target = cmedv)
dim(pdp::boston)
## [1] 506  16

# larger housing data (target = Sale_Price)
dim(AmesHousing::make_ames())
## [1] 2930   81
```

]

```r
# employee attrition data (target = Attrition)
dim(rsample::attrition)
## [1] 1470   31
```

]

*The random forest and boosting module will make use of a few additional data sets 😱

---

background-image: url(Images/yourturn.png)
background-position: center
background-size: contain

# Your Turn!

---

## Meet your neighbors

- Name

- Where you're from

- Experience with R

]
]

]

---

## Download remaining material

---

background-image: url(http://i.imgur.com/9umnm.gif)
background-position: center
background-size: contain

# Excited To Start?