SPOTIFY TURKISH POP MUSIC ANALYSIS

In this final version, this project aims to visualize and define similarities between Turkish Pop music. Also, a Shiny web app wil be provided. Since, the initial versions of the code are shared in the previous Rmd file. The same, API related parts, will not be shared again.

P.S. API requests for the genres excluding Turkish Pop, gathers the data but leaves the artist names as NA if the artists’ genre tags do not match with Pop artists. Thus, when the NA data is cleaned, it is seen that the artists that are remained in the data set, matches with the Pop artists as a subset. For being able to make a substantial analysis, we have decided to exclude the rock data. To show the issue, the NA filtered Rock data is added at the end.

Conclusions For The Final Version

Turkish people loves songs with the content of “Discontent, Uneasiness” followed by “Despair, Wailing and Weeping”
The most positive vibed songs are from the artists: Serdar Ortac and Hande Yener, Karabiberim is the most positive song.
Top5 Artist in the list according to the their occurence in the top50 songs Sezen Aksu (7), followed by Tarkan(6), maNga(5), Teoman(5), Buray(4).
There is no pattern between Turkish Pop hit songs; danceability, energy (activeness, energeticness) and valence (musical positivity) are more common in the hits. It is observed that there is no correlation between the musical attributes
Love is the most common word in the hit songs’ name, in the top songs the remixes have a huge weight.

The key for hit pop songs: It is observed that Turkish people loves the artists with their variated songs. Musical attributes differ for each of their songs. A title with love works well. Remixes also have a chance. Energy, danceability and valence is common between the hits. G minor and C# Minor is the most common tones for the songs, they may be a good choice. Also, the despair and uneasiness is common as a theme.

The libraries:

For this analysis, the libraries that will be used is documented below:

knitr::opts_chunk$set(echo = TRUE)

options(stringsAsFactors=FALSE)
library(tidyverse)
library(lubridate)
library(ggplot2)
library(spotifyr)
library(ggjoy)
library(reshape)
library(treemap)
library(fmsb)
library(vcd)
library(wordcloud)
library(SnowballC)
library(RColorBrewer)
library(tm)
library(reshape2)

load(file ="C:\\Users\\minas\\OneDrive\\Masaüstü\\shinydata.RData")

Explanation of The Data Set - Data Exploration

In this project, we will be using a dataset that includes 193 rows which are the best 20 Turkish Pop Artists’ best 10 songs. The expected dataframe length is 200. Since all of the artists do not have 10 songs the the expected length couldn’t be reached. The table below represents a small set of data with 20 lines.

The data is gathered from Spotify API.

The data includes:

Artist name
Album name
Track name
Artist id
Track id
Track popularity
Album id
Duration in milliseconds
Energy
Key
Loudness
Mode
Speechiness
Acousticness
Instrumentalness
Liveness
Valence
Tempo
Artist popularity
Artist followers

Extras: (The musical attributes that we have received from Spotify ranges from 0-1 in its original version. For ease of use in plots it is rearranged between 0-100)

Danceability will be used between 0-100 (0.0-1.0 in the actual): How likely to dance!
Energy will be used between 0-100 (0.0-1.0 in the actual): If the song is energetic, active or not
Speechiness: “Values above 66 describe tracks that are probably made entirely of spoken words.”
Acousticness: From 0 to 100 decribes whether the track is acoustic
Liveness: “Higher liveness values represent an increased probability that the track was performed live”
Valence: 0-100, describes the musical positivity, cheerfulness etc.

top20Artist_songFeatures <-
  top20_info$top_tracks %>% left_join(.,top20_info$artist_info,by="artist_id") %>%
  select(artist_name, album_name, track_name, everything()) %>%
  mutate_at(vars(danceability,energy,speechiness,acousticness,instrumentalness,liveness,valence),list(~.*100)) %>%
  tbl_df()

knitr:: kable(top20Artist_songFeatures %>% slice(1:20))

artist_name	album_name	track_name	artist_id	track_id	track_popularity	album_id	duration_ms	danceability	energy	key	loudness	mode	speechiness	acousticness	instrumentalness	liveness	valence	tempo	artist_popularity	artist_followers
Tarkan	Ölürüm Sana	Simarik	2yMN0IP20GOaN6q0p0zL5k	6vEL7E8TGzrxuxbrgdlkLL	61	6ebdbUXZn5AURKeRNXqiay	235218	77.5	81.8	7	-8.687	1	16.40	5.180	0.000000	11.40	82.6	97.060	66	1481877
Tarkan	Her Şey Fani (Mahmut Orhan Remix)	Her Şey Fani - Mahmut Orhan Remix	2yMN0IP20GOaN6q0p0zL5k	0JHarUsJHcMWbmmIK8i4mw	58	6D7tNjcwXdmaJgFKrYpIuA	310606	69.0	67.9	7	-7.461	0	3.31	28.100	0.000430	17.00	22.9	113.019	66	1481877
Tarkan	Karma	Ay	2yMN0IP20GOaN6q0p0zL5k	0oVnsgyHclNnGeEb4hRoYO	55	11KfFiHZxXiBoEq52HAsJi	260713	68.5	76.3	11	-6.540	0	8.39	0.443	0.016400	21.70	41.8	90.010	66	1481877
Tarkan	Karma	Kuzu Kuzu	2yMN0IP20GOaN6q0p0zL5k	3ajYmPDtB3P7JTPQ64Dr7P	55	11KfFiHZxXiBoEq52HAsJi	233639	70.5	88.3	4	-8.351	0	5.08	2.230	0.834000	5.26	73.0	93.909	66	1481877
Tarkan	Aacayipsin	Kış Güneşi	2yMN0IP20GOaN6q0p0zL5k	1tiIKelgeYVi1KcylVnZdq	54	06EBiLh3V0q2dOUdXnic7e	235171	65.2	70.6	11	-13.407	0	4.36	1.580	0.000000	8.43	89.2	91.024	66	1481877
Tarkan	Ölürüm Sana	Ölürüm Sana	2yMN0IP20GOaN6q0p0zL5k	6nAgxB5x7vlGVVIoWpCwaM	54	6ebdbUXZn5AURKeRNXqiay	246456	75.3	92.6	9	-8.485	1	10.30	0.718	0.000228	4.07	79.0	92.015	66	1481877
Tarkan	10	Kedi Gibi	2yMN0IP20GOaN6q0p0zL5k	1BfpV5h16V1Gtr41GEF9cO	53	4fJzakARJP2UfOjSj5Q9s1	249006	69.9	89.7	7	-4.477	1	5.85	1.450	0.000405	23.90	56.8	98.028	66	1481877
Tarkan	Karma	Sen Başkasın	2yMN0IP20GOaN6q0p0zL5k	0ZJKPGKFJQM9riXJ8TOyiO	50	11KfFiHZxXiBoEq52HAsJi	256069	72.9	78.8	10	-8.301	0	7.14	3.520	0.003150	9.69	72.2	91.923	66	1481877
Tarkan	Ölürüm Sana	Salina Salina Sinsice	2yMN0IP20GOaN6q0p0zL5k	4PGsEvmwfJwNJJcsw9aaav	50	6ebdbUXZn5AURKeRNXqiay	236054	64.9	80.7	3	-9.757	0	13.50	6.130	0.000505	6.48	76.1	171.997	66	1481877
Tarkan	10	Yolla	2yMN0IP20GOaN6q0p0zL5k	1vZAYNFTcxzrnooQUw5zKD	49	4fJzakARJP2UfOjSj5Q9s1	285856	67.6	91.3	7	-5.251	1	5.37	0.422	0.000225	7.52	65.8	120.003	66	1481877
Sezen Aksu	Bahane	Şanıma İnanma	64d1rUxfizSAOE9UbMnUZd	2aKHw8ukPHrJTwllQBIZSY	56	3bxl6Vvb6z122wXPyvcdmN	235773	70.9	80.0	5	-6.198	0	5.78	28.800	0.000462	14.80	82.8	95.993	68	1998899
Sezen Aksu	Düş Bahçeleri	Kaçın Kurası	64d1rUxfizSAOE9UbMnUZd	3Qaj6EzdVTzysN50rg2u4p	56	3pCSt4dgNRTTDVI1XO6nqZ	225693	60.9	73.1	0	-6.495	0	32.90	45.500	0.038000	29.30	80.8	158.284	68	1998899
Sezen Aksu	Gülümse	Her Şeyi Yak	64d1rUxfizSAOE9UbMnUZd	6JTrpXQBPdEII6C0SbXBYo	56	1m07iv4leo4aoqFs60vHCK	314146	57.0	36.6	2	-8.898	0	3.33	82.400	0.001570	29.60	45.4	106.900	68	1998899
Sezen Aksu	Düş Bahçeleri	Seni Yerler	64d1rUxfizSAOE9UbMnUZd	0seHKl9cX6sRrgO5Ql28b4	55	3pCSt4dgNRTTDVI1XO6nqZ	250306	75.2	81.4	8	-6.787	0	10.30	3.150	0.000264	8.58	76.2	119.026	68	1998899
Sezen Aksu	Sen Ağlama	Tükeneceğiz	64d1rUxfizSAOE9UbMnUZd	2rP7pI2WpMWcUraYAX2xiT	55	13JKU1RyLFS73hDvWnTHr1	259360	39.5	67.5	1	-5.881	0	4.72	64.900	0.000000	9.63	49.6	177.053	68	1998899
Sezen Aksu	Gülümse	Vazgeçtim	64d1rUxfizSAOE9UbMnUZd	6Fov2wtS1YVTS3o4sqhgVM	54	1m07iv4leo4aoqFs60vHCK	309906	48.0	21.7	0	-11.520	0	3.80	78.400	0.076900	11.20	18.8	140.601	68	1998899
Sezen Aksu	DEMO	Begonvil	64d1rUxfizSAOE9UbMnUZd	1JYxCgv4Jlx2X4SYNtXgkB	54	6yYVJKYaBp37t9zJic1dA2	245835	45.0	27.6	8	-5.900	0	4.01	89.200	0.000000	12.90	35.0	89.297	68	1998899
Sezen Aksu	Öptüm	Vay	64d1rUxfizSAOE9UbMnUZd	64wJJQ7wFtopVHVU3QWVxk	53	1hOFeE5hZYxIO72UKhnrWS	237540	40.4	42.7	11	-9.561	0	3.89	65.900	0.000000	11.30	44.0	120.172	68	1998899
Sezen Aksu	Gülümse	Seni Kimler Aldı	64d1rUxfizSAOE9UbMnUZd	0o2xtLCzHseeiUMYMGFDcC	53	1m07iv4leo4aoqFs60vHCK	208546	27.2	36.5	9	-8.257	0	2.89	77.400	0.000000	10.90	16.3	92.035	68	1998899
Sezen Aksu	DEMO	Anlasana	64d1rUxfizSAOE9UbMnUZd	48nXZHMaiXLQP9KTVoh9Bj	52	6yYVJKYaBp37t9zJic1dA2	197879	70.3	64.6	10	-7.965	0	13.60	1.660	0.007570	26.30	59.9	110.952	68	1998899

We had fun :)

We looked for the most positive songs in the top20 artists’ best songs. And it was “Karabiberim” from Serdar Ortac which is also our favorite!!!

Musical_Positivity <-
  top20Artist_songFeatures %>%
  arrange(desc(valence)) %>% 
  select(artist_name,track_name, valence) %>% 
  head(10)

Musical_Positivity %>%
  knitr::kable()

artist_name	track_name	valence
Serdar Ortaç	Karabiberim	96.8
Hande Yener	Aşkın Ateşi	96.1
Hande Yener	Mor	96.1
Sıla	Kafa	96.0
Buray	Tac Mahal	95.2
Hande Yener	Kırmızı	94.5
Gülşen	Yukardan Ayarlı	94.3
Serdar Ortaç	Gamzelim	94.2
Mustafa Sandal	Jest Oldu	93.5
Demet Akalın	Afedersin	92.6

Joyplot for Musical Positivity (Valence)

The plot below represents the visualization of the artists in top20 accordingly with their songs’ valence. Valence is considered as a key parameter that determines whether the music has positive vibes or not.

Since the plots are more skewed to 100 for Serdar Ortac, Buray, Hande Yener, Tarkan, Kenan Dogulu and Mustafa Sandal, it is fair to say that they have more positive songs compared with the other artists that are included in the data set.

ggplot(top20Artist_songFeatures, aes(x = valence, y = artist_name)) + 
  geom_joy() + 
  theme_joy() +
  ggtitle("Joyplot of Top Artists' Musical Positivity", subtitle = "Based on valence / measurement of positivity")

The Top 10 Turkish Pop Songs - How Their Attributes Match?

The bar graph below represents the musical attributes for the best 10 songs accordingly with their track popularity.

It is seen that danceability and energy is high and valence is over the average for the best 10 songs.

knitr::kable(Top10_songs)

artist_name	album_name	track_name	artist_id	track_id	track_popularity	album_id	duration_ms	danceability	energy	key	loudness	mode	speechiness	acousticness	instrumentalness	liveness	valence	tempo	artist_popularity	artist_followers
Aleyna Tilki	Nasılsın Aşkta?	Nasılsın Aşkta?	4ckLjJztj53Ifid7WHweBn	4Am7XgKpt84IDRCv3PyXW8	70	545QgjpFzjjmCTpuj7zcAZ	171080	71.0	78.2	8	-6.703	0	4.66	11.800	0.000192	10.60	21.6	139.932	62	416575
Gülşen	Bir İhtimal Biliyorum	Bir İhtimal Biliyorum	1F2v33FQTavJyaD7ZsyhdA	24mpJgP2YZUBif9NvmYyh4	67	2CICALHFSoebVWZqi9hvUo	210232	69.2	51.8	1	-9.391	1	4.07	1.830	0.000000	28.40	32.6	98.020	62	619329
Mustafa Sandal	Gel Bana	Gel Bana	0mkH5jj3goQ51JtPKVodTo	3AXScNtH6Ywywwma6u8h95	65	2yRDMGRoahDlyRWfmhGT2Q	183337	86.7	58.3	4	-8.782	0	35.90	18.000	0.005050	14.70	45.1	90.017	62	213921
Yalın	Deva Bize Sevişler	Deva Bize Sevişler	46zuW8tHxwahYn7VNMgYTa	7l8e7Pe8qndxhCU245rIve	65	4L1PIFhDNf4qc1K2yBQqOh	218740	52.8	81.1	8	-7.147	0	6.51	30.400	0.002210	5.65	74.8	166.022	64	415726
Buray	Kehanet	Kabahat Bende	1qZ684TB9E1BjH58btdtYd	3UgBLWFgadLahNGxTDYG2d	65	54fUFXduCOTJEksa3bUU89	206700	57.2	71.8	11	-5.771	0	5.56	41.300	0.000000	11.20	87.7	79.839	63	466051
Hande Yener	Beni Sev	Beni Sev	08mjMUUjyTchMHCW7evc3R	0bE8X3KwUifITMaibVR9eD	63	2xUTDtK1rGCdFmEqUKFCLw	259093	58.5	76.6	2	-4.750	0	3.29	0.678	0.001360	42.20	36.3	110.006	61	418609
Simge	Ben Bazen	Öpücem	4StjyzjcmZ7a9QncHVf0pu	1pq9w58vpi5nnEllpgl1tF	62	7JkYuwCd3THEGv2YZYgSDE	255150	74.6	80.2	8	-6.590	0	5.75	6.070	0.000929	14.30	81.6	132.010	61	311886
Tarkan	Ölürüm Sana	Simarik	2yMN0IP20GOaN6q0p0zL5k	6vEL7E8TGzrxuxbrgdlkLL	61	6ebdbUXZn5AURKeRNXqiay	235218	77.5	81.8	7	-8.687	1	16.40	5.180	0.000000	11.40	82.6	97.060	66	1481877
Sıla	Acı	Sabır	5gFPi3KWXEwA9bLEO47Ow0	6I6gNA8Slf7kdSQZ9kObe3	61	33FUvw367XssYPX3LcWiFd	195886	74.9	89.3	9	-5.612	0	8.55	52.500	0.000000	13.00	77.9	120.021	62	1127473
Demet Akalın	Ateş	N’apıyorsan Yap	1U449OOb70EZlElNjLMwCM	7w8XcjYxuykXdg9PTDTjBR	60	0WAS8LZqjH2yhUbFyEkDtI	173000	72.1	74.8	5	-8.564	0	13.20	2.180	0.000000	7.07	62.5	170.054	64	690833

ggplot(Top10_songs_v1.long, aes(x=variable, y=value))+geom_bar(aes(y=value, fill=track_name),stat="identity", alpha=0.8 , position="dodge")+ ylab("Value")+ xlab("Features")+coord_flip()+ggtitle("Top 10 Turkish Pop Songs in Spotify 2019")

Who knows best to be in the Top50 :)

It is observed that the most popular artist “Sezen Aksu” has the most songs in the list. Interestingly, even Aleyna Tilki has only 3 songs that are available in the market, she has achieved the success to be in the top 50 with 2 of them. Also, Emrah Karaduman has songs with her and his most popular songs with Aleyna Tilki.

As a keypoint, if you want to be in the hits, be like Aleyna Tilki

Top50%>%count(artist_name,sort=TRUE)

## # A tibble: 15 x 2
##    artist_name         n
##    <chr>           <int>
##  1 Sezen Aksu          7
##  2 Tarkan              6
##  3 maNga               5
##  4 Teoman              5
##  5 Buray               4
##  6 Emrah Karaduman     4
##  7 Yalın               4
##  8 Demet Akalın        3
##  9 Simge               3
## 10 Aleyna Tilki        2
## 11 Deeperise           2
## 12 Sıla                2
## 13 Gülşen              1
## 14 Hande Yener         1
## 15 Mustafa Sandal      1

What about statistics?

We have searched for correlation between the musical attributes of the 193 tracks and observed that there is no correlation between the musical attributes. Sadly, we are not able to come up with a conclusion like: “If the song is energetic, you will be dancing”.

correl_values=top20Artist_songFeatures %>%
  select(danceability,energy,speechiness,acousticness,instrumentalness,liveness,valence)

correl_values <- cor(correl_values)

correl_values %>%
  knitr::kable()

	danceability	energy	speechiness	acousticness	instrumentalness	liveness	valence
danceability	1.0000000	0.3821162	0.1682726	-0.4695240	-0.0317080	-0.1061904	0.4622876
energy	0.3821162	1.0000000	0.2182280	-0.6810877	-0.3933557	0.0836172	0.5809970
speechiness	0.1682726	0.2182280	1.0000000	-0.1714036	-0.1308977	0.0633878	0.1661581
acousticness	-0.4695240	-0.6810877	-0.1714036	1.0000000	0.2210678	-0.0526673	-0.3319874
instrumentalness	-0.0317080	-0.3933557	-0.1308977	0.2210678	1.0000000	-0.1296594	-0.2418518
liveness	-0.1061904	0.0836172	0.0633878	-0.0526673	-0.1296594	1.0000000	-0.1250374
valence	0.4622876	0.5809970	0.1661581	-0.3319874	-0.2418518	-0.1250374	1.0000000

Snapshot for Track Attributes for the Most Listed Artists

With the radar charts below, we have concluded that the best songs attributes variate a lot even in the same artists’ songs. There is no pattern within the artists own songs or with the other artists. So, they are jack of all traits, also the master of them.

plotArtistRadar(theArtist = "Sezen Aksu",Top50)

plotArtistRadar(theArtist = "Tarkan",Top50)

plotArtistRadar(theArtist = "maNga",Top50)

plotArtistRadar(theArtist = "Teoman",Top50)

plotArtistRadar(theArtist = "Buray",Top50)

# Curiosity:
plotArtistRadar(theArtist = "Aleyna Tilki",Top50)

plotArtistRadar(theArtist = "Emrah Karaduman",Top50)

What about emotions, what does Turkish people love listening to?

Key Characteristics and Mood Link

C major :Innocently Happy
C minor :Innocently Sad, Love-Sick
C sharp minor : Despair, Wailing, Weeping
C sharp major: fullness, sonorousness, euphony
D major: Triumphant, Victorious War-Cries
D minor: Serious, Pious, Ruminating
D sharp minor: Deep Distress, Existential Angst
D sharp major: Cruel, Hard, Yet Full of Devotion
E major: Quarrelsome, Boisterous, Incomplete Pleasure
E minor: Effeminate, Amorous, Restless
F major: Complaisance and calm
E sharp minor: Obscure, Plaintive, Funereal
F sharp major : Conquering Difficulties, Sighs of Relief
F sharp minor: Gloomy, Passionate Resentment
G major: Serious, Magnificent, Fantasy
G minor: Discontent, Uneasiness
G sharp major : Death, Eternity, Judgement
G sharp minor : Grumbling, Moaning, Wailing
A major : Joyful, Pastoral, Declaration of Love
A minor : Tender, Plaintive, Pious
A sharp major: Joyful, Quaint, Cheerful
B flat minor : Terrible, the Night, Mocking
B major: Harsh, Strong, Wild, Rage
B minor: Solitary, Melancholic, Patience

According to the top 50 songs mode and key distribution, we have observed that Turkish people loves songs with the content of “Discontent, Uneasiness” followed by “Despair, Wailing and Weeping”

Top50$tone <- ifelse(Top50$mode==0, "minor", "major")

Top50$scale <-

                      ifelse (Top50$key==0, "C",
                      ifelse(Top50$key==1, "C#",
                      ifelse(Top50$key==2, "D",
                      ifelse(Top50$key==3,"D#" ,
                      ifelse(Top50$key==4, "E",
                      ifelse(Top50$key==5, "E#",
                      ifelse(Top50$key==6,"F",
                      ifelse(Top50$key==7, "F#",
                      ifelse(Top50$key==8, "G",
                      ifelse(Top50$key==9,"G#",
                      ifelse(Top50$key==10,"A","A#"))))))))))) 


Top50$keys <- paste(Top50$scale, Top50$tone, sep= " ")


tone1 <- group_by(Top50, keys)
tone2 <- dplyr::summarise(tone1,  count=n())
tone2 <- arrange(tone2, desc(count))

# Tonality treemap
treemap(tone2, index="keys", vSize="count", type="index", 
        palette="Pastel1", title="Top 50 Songs Key characteristics", fontsize.title=12)

# major vs Minor
major <- group_by(Top50, tone)
major2 <-dplyr::summarise(major, count=n())

# Major treemap
treemap(major2, index="tone", vSize="count", type="index", 
        palette="Pastel2", title="Top 50 Songs Major vs Minor", fontsize.title=12)

Turkish Pop Hits’ Names!

In the best 193 tracks of 20 best artists, the most common word that is included in the hits names is “Love”
Also, the remixes are common between the songs

names <- gsub("remix", "", names, ignore.case = TRUE, perl = FALSE,
    fixed = FALSE, useBytes = FALSE)
wordcloud(names, max.words=20 ,random.order=FALSE,rot.per=0.35,colors=brewer.pal(5, "Dark2"), main="Title")

# A total match as a subset of Turkish pop tagged artist with also the rock tag
collectionAll_filtered %>%
  filter(Genre=="turkish rock")

## # A tibble: 70 x 22
##    artist_name album_name track_name artist_id track_id track_popularity
##    <chr>       <chr>      <chr>      <chr>     <chr>               <int>
##  1 Tarkan      Ölürüm Sa~ Simarik    2yMN0IP2~ 6vEL7E8~               61
##  2 Tarkan      Her Şey F~ Her Şey F~ 2yMN0IP2~ 0JHarUs~               58
##  3 Tarkan      Karma      Ay         2yMN0IP2~ 0oVnsgy~               55
##  4 Tarkan      Karma      Kuzu Kuzu  2yMN0IP2~ 3ajYmPD~               55
##  5 Tarkan      Aacayipsin Kış Güneşi 2yMN0IP2~ 1tiIKel~               54
##  6 Tarkan      Ölürüm Sa~ Ölürüm Sa~ 2yMN0IP2~ 6nAgxB5~               54
##  7 Tarkan      10         Kedi Gibi  2yMN0IP2~ 1BfpV5h~               53
##  8 Tarkan      Karma      Sen Başka~ 2yMN0IP2~ 0ZJKPGK~               50
##  9 Tarkan      Ölürüm Sa~ Salina Sa~ 2yMN0IP2~ 4PGsEvm~               50
## 10 Tarkan      10         Yolla      2yMN0IP2~ 1vZAYNF~               49
## # ... with 60 more rows, and 16 more variables: album_id <chr>,
## #   duration_ms <int>, danceability <dbl>, energy <dbl>, key <int>,
## #   loudness <dbl>, mode <int>, speechiness <dbl>, acousticness <dbl>,
## #   instrumentalness <dbl>, liveness <dbl>, valence <dbl>, tempo <dbl>,
## #   artist_popularity <int>, artist_followers <int>, Genre <chr>

SPOTIFY TURKISH POP MUSIC ANALYSIS

Mina Silahtaroglu - Umre Metin

17 May 2019