Visualising my music catalog with matplotlib

I have always wondered how my music catalogue would look like if visualised. Mainly wondered because when shuffling the music the same type of music got played. But I know I tried to make my music choices to more diverse. so I decided to cook up some code to analyse my music library.

First, I need a way to get the music in a table like a format, So I googled around and found that you can use this command dir > filenames.txt to quickly export filenames to a CSV file. But the exported file has a lot of info that was not needed for my project. Then I cleaned the data up by deleting the all the columns expect filenames. But instead, I opted for a different method as having a program called mp3tag which helps you add details to your music files like artists, album info and most importantly for me added album pictures to the file. The program has an option to export the files to a CSV file format. This file was cleaner with more relevant data like album or artist.

Which can be used for more use in the project if needed. Not just filenames. The main detail it lacked was the genre of the songs so I had to manually add them myself. this did not take to long maybe around a few hours.

After that, I decided to start coding the project. I based a good amount of my code on this medium post, as I did not how to use pie charts in matplotlib. First, I used pandas to read the CSV file to the script and the data massaging

Using after reading the CSV to the file I previewed the data with the head() function. Looking at the data I found duplicate genres from simple misspellings. So I fixed by going directly to the CSV file and editing out the issues. Next, I used the value_counts() function to print the frequency of each genre, from this I quickly found my music list is less diverse than I thought. Now, this needed to convert the data frames to a format that the matplotlib pie function to understand. To do this I converted the value_counts to lists. And separated by name of genres and the number of counts. This was done by using the .valuecounts().index().tolist() function the index()is used so you can extract the names so you have a list like this ["Kpop", "pop", "Rap", "Film"]To get the list with the numbers only you can omit the index()like this .value_counts().tolist()

Now you can print the variables to see if everything is all good. So you can move on to plotting the data. First, we define the colours of our pie chart. I picked the colours using a hex picker from google to did the colour scheme a minimal style look. Next, I turned the lists from earlier to numpy arrays so to can be used for the legend. This is because i had technical difficulties moving around the text of the percentages to prevent clumping up so found this solution to use a legend instead thanks to this StackOverflow article.
We make percentage variable and plot legends below:

percent = 100.*y/y.sum()
percent2 = 100.*y2/y2.sum()
patches, texts, = ax1.pie(y, colors=colors, startangle=90, radius=1.2)
patches2, texts2 = ax2.pie(y2, colors=colors, startangle=90, radius=1.2)
labels = [‘{0} — {1:1.2f} %’.format(i,j) for i,j in zip(x, percent)]
labels2 = [‘{0} — {1:1.2f} %’.format(i,j) for i,j in zip(x2, percent2)]

The labels are for the legend showing the percentage with the genre. 
We apply another style charge to the pie charts by changing the text to grey:

for text in texts: text.set_color(‘grey’)

This is the code that plots the legends to the figures:

for text in texts: text.set_color(‘grey’)

This is the code that plots the legends to the figures:

sort_legend = True    

if sort_legend: patches, labels, dummy = zip(*sorted(zip(patches, labels, y), key=lambda x: x[2], reverse=True))]

patches2, labels12, dummy2 = zip(*sorted(zip(patches2, labels2, y2), key=lambda x: x[2], reverse=True))

ax1.legend(patches, labels, loc=’center left’, bbox_to_anchor=(-0.1, 1.), fontsize=8)

ax2.legend(patches, labels2, loc=’center right’, bbox_to_anchor=(-0.1, 1.), fontsize=8)

Then you should have some pie charts that look something like this:

My Machine Learning Challenge

 

This is a learning project I want to do for a while. The learning material will be based on Siraj Raval’s Learn Machine Learning in 3 Months video. He gives a clear road to becoming competent in machine learning. Have been playing with some of the coding challenges from the learn python for data science section so I won’t be doing this blind.

The first month is mainly math and theory-based, Before taking the MIT linear algebra course will be taking 3Blue1Brown’s Essence of linear algebra first to develop an intuition for the topic. So I don’t listen to the lectures with no clue of whats going on.

To be honest, I think can finish this project in less than 3 months as my summer holidays just started and I will be spending most of my time on this project. But I do get described easily by other learning projects, for example, I want to start an Instagram page to practice my photography and marketing skills(Growth Hacking) and improve my guitar skills or learn some new sports or hobbies. Too ambitious? :)

I will be trying to be blogging about this project every day to keep me accountable and also help improve my writing. I got this inspiration from Praxis 30 day blogging Challenge.