UnsupervisedLearning.Rmd

---
title: "Unsupervised Learning"
output: html_notebook
---

Two different goals:
* Looking for patterns in data where no outcome is defined (groups of observations)
* Looking for related features (groups of variables)

# K-means
# Hierarchical clustering
# PCA
## Intro
Principal components are linear combinations of variables
Retains/explains the maximum variation for a given number of components
   Regression line is the first principal component
They are orthogonal to each other

Issues;
* Scaling
* Missings
* categorical variables

[Luke Hayden article on PCA on Datacamp](https://www.datacamp.com/community/tutorials/pca-analysis-r)

```{r}
library(tidyverse)
library(ggbiplot)
```

```{r}
mtcars_pca <- prcomp(mtcars, center = TRUE, scale = TRUE)
 
ggbiplot(mtcars_pca) + theme_light()
```

```{r}
iris_pca <- prcomp(iris[1:4], center = TRUE, scale = TRUE)

ggbiplot(iris_pca, groups = iris$Species) + theme_light()
```
```{r}
UKeats <- tibble(
foods  = c("Alcoholic drinks", "Beverages", "Carcase meat", "Cereals", "Cheese", "Confectionery", "Fats and oils", "Fish", "Fresh fruit", "Fresh potatoes", "Fresh Veg", "Other meat", "Other Veg", "Processed potatoes", "Processed Veg", "Soft inks", "Sugars"),
England = c(375,57,245,1472,105,54,193,147,1102,720,253,685,488,198,360,1374,156),
NIreland = c(135,47,267,1494,66,41,209,93,674,1033,143,586,355,187,334,1506,139),
Scotland = c(458,53,242,1462,103,62,184,122,957,566,171,750,418,220,337,1572,147),
Wales = c(475,73,227,1582,103,64,235,160,1137,874,265,803,570,203,365,1256,175)
)

UKeats_pca<- prcomp(UKeats[2:5], center = TRUE, scale = TRUE)
summary(UKeats_pca)

ggbiplot(UKeats_pca) + theme_light()
```

```{r}
UKeats2 <- UKeats %>% gather(country, quantity, 2:5) %>% spread(foods, quantity)
UKeats2_pca<- prcomp(UKeats2[2:18], center = TRUE)
UKeats2_pca
summary(UKeats2_pca)

ggbiplot(UKeats2_pca, labels =UKeats2$country) + theme_light()
```