Regression
This analysis focuses on the relationship between time and loudness.
This part of regression mainly involve running on Spotify Data.
Is the loudness for entire music getting louder with the time goes?
The “loudness war” is very interesting in music history. It happened by the psychoacoustic phenomenon, which reflects the mentality of people preferring louder music. This happens in the past four decades and raised an extreme competition in music loudness. Based on this historical phenomenon, we consider if the linear regression model can demonstrate what it happened in this period.
Data
The datasets we used is the songs data. We uses attribute ‘release_date’ and ‘loudness’. We change ‘release_date’ as integer year such as ‘1999’ and ‘2008’ from its original format datetime( year-month-day ) . Therefore, we get a series of years and the ‘loudness’ of each sound in various years. Some years may include very large amount of ‘loudness’ feature, but some have few or none, which is shown as figure1.
In figure 1, it is clear that the most of the songs are in the period of 1950 to 2020. Besides, we can see that with the year increases, the range of loudness expand. Maybe it caused by the reason that music becomes more diverse. just like we mentioned that there were more experiment music created and people became much more innovative with the time increase. Thus, we make a new train datasets which include only mean of loudness for each year, and the result shown as figure2. Besides, because of lack of sufficient data, the points before 1960 are very discrete. But the hypotheses is mostly concern about the data from 1980 to 2020. even without the models we can obviously find out the data has trend of increasing in these years.
Model
We create polynomial regression model and SVM model(SVR) in regression to figure out the trend of loudness by the year pass. In polynomial regression model, we chose the degree 5 in the model to figure out the tendency of music loudness.
The reason why we choose these 2 model as regression model in this hypotheses is that:
-
Choosing the polynomial regression is to do the further approach of linear regression model, and see the details of the trend changing.
-
SVM regression can tell the details of changing and the trend in a small period.
From the 3 figure above, the result of polynomial regression demonstrates the how much the increase change. Interestingly, using the mean value or not show to different result of increasing change, this may caused by the issue of whether the data is sufficient. Then, the SVM regression shows the most details. We can see how sharp the loudness increase after 1990, which may be the time when the fiercest competition happens. And we can see the loudness remains at -10 after 2000s. Maybe the loudness reach it highest level that it can reach or maybe people’s mind change. These guess may not be true, but the results do shows the “loudness war” happens during the time after 1980. The music are louder than before.