Today  we  would  like  to  present

Relationship  between  the  Type  of  Music and  Economic  Conditions.

This  is  our  agenda  for  the  presentation.

Firstly,  we  will  introduce  our  data set

and  then  see  the  meaning of  variables  as  dictionary.

Third,  we  will  talk  about our  hypothesis  for  our  study.

Fourth,  we  will  go  through our  background

and  then  the  data  processing  part

such  as  changing  categorical  variables to  continuous  variables

and  handling  our  missing  values and  outliers  and  then  split  the  data set.

Next,  we  will  discuss the  best  model  in  our   dataset.

Lastly,  we  will  give  business  values  and recommendations.

In  this  study,  we  will  analyze the  song's  popularity  on  Spotify  in  US

from  2006  to  2019 as  well  as  economic  indicators

such  as  GDP,  inflation r ate, unemployment  rate,  and  happiness  variable.

In  this   dataset,  we  have 1,389  rows  and  20  columns.

We  have  taken  this   dataset  from  Kaggle.

Furthermore,  other   datasets are Morris for  economic  indicator.

We  can  see  more  detailed  information on  the  next  page.

This  is  our  dictionary  of   dataset.

The  target  variable  is  popularity

which  means  the  higher  the  value, the  more  popular  the  song  is.

It  would  be  the  continuous  variable.

In  the   predictor variable, we  have  musical  attributes

such  as  explicit, danceability,  and  loudness.

Also,  we  have  four  variables for  economic  indicators

which  are  GDP,  happiness  index, unemployment  rate,  and  inflation.

Our  hypothesis  is  that  during  times of  economic  downturn,

we  suppose  there  is  an  increase in  the  popularity  of  upbeat

and  energetic  genres such  as  electronic  dance  music,

while during  economic  growth  periods,

more  soulful,  introspective, thoughtful  music

tend  to  dominate  the  charts.

We  expect  that  it  provides valuable  insights  into  the  culture

and  economic  factors that  influence  the  music  industry

and  can  assist  music  producers and  marketers

in  understanding  consumer  behavior and  making  informed  decisions.

We  are  using  2006 to  2019  year  in  the  US.

It  shows  that  2006  to  2007 is  pre-recession  time,

so economy  was  still stable.

2008  to  2009  went  through a global  financial  crisis.

2010  to  2012 is  post- recession  recovery  time.

In  this  time, the  economy  began  recovering.

In  2013 to  2015  is  steady  recovery  time.

In  2016  to  2017, the  economy  continued  to  grow.

Finally,  2018  to  2019  is late- stage  growth  time.

In  this  time,  it  was  approaching a  more  mature  and  stable  state.

We  found  that  the  worse  the  economy,

the  more  positive the  type  of  music  people  listen  to.

Based  on  this  timeline,  we  can  get a  correlation  with  other  variables.

As  you  can  see  in  this  graph, we  can  see  that  unemployment  rate

is getting  higher  during  2008  to  2012 which  is  a  global  financial  crisis.

Based  on  this  hypothesis,

positive  music will  be  popular  for  this  time.

This  is   about the tempos  and  the  years.

The  lower  value  means  the  lower  tempos and  higher  value  means  the  higher  tempos

which  we thinks  that  lower  tempos

may  appear  in  the  sad  song or   softer  songs

and  the  higher  tempos  may  appear in  the  inspiring  songs  or  happier  songs.

We  look  into  2006  through  2007 when  the  economy  was  growing  and  stable,

you  could  see  that  they  accounted… The  lower   tempo have  more…

Lower- tempo  songs  appeared more  parts  in  this  period.

If  we  look  into  2013  through  2015 and  2016  through  2017

when  the  economy  was  growing, we  could  see  that  they  also

have  more  lower- tempo  songs compared  with  the  other  years.

If  we  look  into  2008  to  2009 and  2010  through  2012

when  the  economy  was  really  bad,

2008  through  2009 was the  financial  crisis

and  2010  through  2012, the  unemployment  rate  was  the  highest.

You  can  see  that  higher- tempo  songs have  more  parts  in  those  years.

We  got  a  conclusion  that  people might  turn  to  listen  to  higher- tempo  songs

when  the  economy  was  really  bad

and  they  re turned to  listen  to  the  lower  songs

when  the  economy  was  good.

Next  page.

We  look  into  valence  and  the  years.

The  lower  valence means  that  it  will  be  a  negative  songs

or  a  lot  of  softer  songs,

and  higher  values  means that  it  will  be  happier  songs,

positive  songs.

We  look  into  2008  through  2009 and  2010  through  2012,

economic  bad  period.

We  could  see  that  the  positive  songs appeared  more  parts  in  those  periods.

In  2013  through  2015 and  2016  through  2017,

the  economy  was  growing,

they  accounted  more  part in  the  negative  songs

which  match  our  hypothesis, seems  that  people  might  listen

to  more  negative  songs when  the  economy  was  good

and  people  want  to  listen to  more  inspiring  songs  or  positive  songs

when  the  economy  was  really  bad.

Next  Page.

Look  into  the  total  number  of  hit  songs of  each  years.

We  can  see  that  through  2013  to  2017 the  hip  hop  songs  was  increasing

and  2018  to  2019,

there  have  more  genre  songs appear  in  this  period.

The  hip  hop  songs  and  the  pop  song and  the  dance  songs  decreased,

but  the   Latin song and  the  R &B  songs  increased.

Next  page.

It  will  be  clear  if  we  look into  this  graph,

we  could  see  that  the  other  genre of  songs accounted  more  parts  in  2018  through  2019.

If  we  look  through  the  time  period,

we  could  see  that  while  the  economy was  growing,  the  other  genres of songs

accounted  more  parts than  the  economic  was  really  bad.

Next  page.

When  we  compare  with  the  valence and  the  GDP,

we  may  find  that  the  valence  and  the  GDP are  negatively  correlated  to  each  other.

It  seems  that  when  the  GDP  was  high

we  will  say  that  the  economy  was  good and  the  valence  will  be  lower.

It  means  that  people  may  turn  to  listen to  the  negative  songs

when  the  economy  was  really  good,

but  they  will  listen  to  the  positive  songs when  the  economy  was  really  bad.

Next page.

Before  we  do  our  model,

we  look  into  our   dataset and  we  found t hat the genre

was  the  only  categorical  variable.

We  transform  it  into  seven  columns of  binary  variables.

It  will  be  blues,  classical,  country dance,  easy  listening,  folk,  and  hip  hop.

Next  page.

We  also  deal  with  the  missing  value.

We  found   [inaudible 00:10:38]  r ows. We  excluded  all  of  them.

The  next,  we  looked  at  the  outliers and  their  distribution.

It  turns  out  the  only  column, instrumentalness,  had  86  outliers.

We  decided  to  transform  the  column

using  SHASH  transform to  normalize  the  column

and  on  the  right  side, you  can  see  that  it's  normalized.

Next  page,  please.

After  cleaning  the   dataset, now  it  is  time  to  split  the   dataset.

We  divided  0.6, 0.2,  and  0.2 to  training,  validation,  and  test   dataset.

Next  slide,  please.

After  applying  to  several  different  models such  as  decision  tree,  regression,

[inaudible 00:11:46] ,  KNN,  neural  network, we  found  that  the  decision  tree

has  the  highest  R Square

and  it  proves  that  it  failed  to  reject our  null  hypothesis.

Next  slide,  please.

When  you  look  close to  the  column  contributions,

other  than  the  music  factors such  as  duration,   acousticness,  key,

we  also  found  that  unemployment  rate,

GDP,  and  inflation  rate also  has  positive  correlation

with  our  target  variable,  popularity.

But  among  the  economic  factors, the  unemployment  rate

has  the  highest  impact on  the  target  variable.

But  other  than  that,

the  music  factors  such  as  duration, acousticness,  key,  loudness,  and  valence

are  the  top  five  contribution to  our  target  variable.

Next  slide,  please.

Based  on  our  model's  result, we  can  say  that  unemployment  rate

has  the  highest  correlation among  economic  factors  with  popularity,

but  also  other  factors have  positive  correlation

with  the  target  variable.

As  unemployment  rate and  [inaudible 00:13:15]   rate  gets  higher,

the  songs  that  has  high  duration, acousticness,  key,  loudness,

valence,  speechiness, energy,  and  danceability  are  popular.

Basically,  the  song  that  is  longer, more  brighter,  and  more  danceability

has the  higher  popularity during  those  times.

Next  slide,  please.

According  to  our  study, we  found  that  the  economic  impact

of  the  popularity of  different  types  of  songs

so  we  recommend  that  during the  economic  turn down,

we  recommend  that  the  music  producers to  make  high- duration  acousticness,  key,

loudness,  valence,  speechiness,  energy, and  danceability  music  and  vice  versa.

Meanwhile,  when  the   economy is  booming,

we recommend  the  music  producers to  make  low  duration,

acousticness,  key,  loudness, valence,  speechiness,

energy,  and  danceability  music.

Next  slide,  please.

This  is  our  reference.

This  is  pretty  much  everything we prepare  for  this  presentation.

Thank  you  so  much  for  listening.