songclst.rmd

---
title: "Untitled"
author: "Jesper"
date: "3/8/2020"
output: html_document
---
### Song clustering

```{r song, echo=FALSE}
rush_song_features <- rush %>% 
  filter(is_studio) %>% 
  select(album_name, track_name, album_release_date, danceability, energy, key, loudness, mode, speechiness, acousticness, instrumentalness, liveness, valence, tempo)

fviz_dist(
  get_dist(
    rush_song_features[,c("danceability", "speechiness", "valence", "energy", "acousticness", "liveness")], 
    method="pearson"), 
  order=TRUE, 
  gradient = list(low = "green", mid="white", high = "red")
  ) + 
ggtitle("Distance matrix of Rush albums (pearson distance metric)") +
theme_minimal()

```

***
Comment

### Clustering on songs

```{r, echo=FALSE}
rush_song_features.pr <- prcomp(rush_song_features[,c("danceability", "speechiness", "valence", "energy", "acousticness", "liveness")])

rownames(rush_song_features) <- rush_song_features[,2]


rush_song_features.cluster <-
  kmeans(
    rush_song_features.pr$x[,c(1:4)],
    centers = 6
  )

plt <- fviz_cluster(rush_song_features.cluster, data=rush_song_features[,c("danceability", "speechiness", "valence", "energy", "acousticness", "liveness")])
plt
```

***

Comment

### Album distribution over clusters

```{r, echo=FALSE}
rush_song_features$cluster <- rush_song_features.cluster$cluster
rush_song_features %>% 
  ggplot(aes(album_name)) +
  geom_histogram(stat="count") +
  facet_wrap(~cluster, ncol=3) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))
```