Knowing yourself better

Recently I came across a video where one IIT professor was reacting to his student’s video. In that video, the student said that he took Mechanical Engineering just because he was good in Maths and Physics, and at the same time he was also advocating why Software or any IT job might be a better option for an average student. I had a spontaneous smirk when he said that thing about Maths and Physics. This was the reason for me for taking Mechanical as well. Nowadays, whenever anyone asks me what is that one mistake that I would like to correct, I always say I would undo this one.

I used to enjoy studying a lot. Back in 10+2, solving Math problems were like a second hobby for me, after, obviously, watching football matches. But college life changed everything. A thing of pleasure started feeling like a huge burden. My college routine piled up more misery on top of that. And things at times seemed unbearable.

Except for a few topics like Thermodynamics, Heat Transfer, etc. (which were ‘Math heavy’), I hardly enjoyed any subject. Then I came to know about Operations Research (OR) in my 4th year. That subject was a kind of revelation for me. Whenever I used to do Math, I would have one question in my mind: How can this be directly applied to real-life problems? OR kind of made me realize on a large scale how we can use Mathematics for that.

Looking back at my college life and those ‘sad boi’ hours, I kind of feel contented that I went through this. You know there is a pattern in sitcoms where they will introduce a side character to bring together the main couple of the show (Karen in The Office, Kelly in Superstore, etc.). My college life kinda did that for me. I realized what I should be doing. Where I am good at. I stopped pursuing MS in Mechanical Engineering and instead focused on doing something related to Math or Data. Data, numbers, stats have been something that always made sense to me. And then I decided to do my Master’s in OR. Now I use OR techniques on a daily basis to make a positive impact for my organization.

Apart from a long retrospection on a lazy Saturday afternoon, there is another reason for writing this blog post. It irks me whenever any of my juniors ask me whether he/she should switch to Data Science to start earning more. If your intention to switch is just to earn more, then best of luck! After some hardships, I understood that I would rather fail in doing something I like. And I feel, for everyone, it is important to find that ‘something’. When you are not employed or associated with any institution, it is important to introspect. The focus should be on knowing yourself better, not mindlessly running after every job opportunity that is available. Otherwise, those ‘sad boi’ hours in life will just be prolonged.

I have been listening: Analysing my stats is a great website for anyone who listens to music (which I guess is everyone) and loves Data. I have been using the site to track what I have been listening since May 2020. So after almost using the site for a year, I decided to decode how I have been listening. This article is mostly focused on the question of ‘how’ rather than ‘what’ because the website pretty much tells you what any user listens. If anyone is interested in extracting any insight of their listening habit then refer to the link provided below. I am sharing the link for my R Script file at the bottom. Only thing you have to do is to download the data from this website. The timeframe considered for this study is June 2020 to May 2021.

Distribution of songs per day:

As it can be seen the distribution for number of songs listened to per day is right skewed. A right skewed data generally suggests that the lower bounds are very low with respect to the whole data. For this case, it will be translated into something like this: There have been many days when I have listened to very few songs. Very few here signifies a number that is very less compared to the daily mean and medium.

Distribution of songs per month:

This graph is also right skewed but it has also an added feature: it is multimodal. We have one peak that is larger than the other two. The other two are similar. This explains that the number of Scrobbles per month has 2-3 groups. One group where the number might be high, one medium, and one low. These high, medium and low distinctions are obviously with respect to the whole dataset. It can be summarised better if we take a look at the number of songs per month for one year:

As it is evident: the number has varied hugely. From 2000 in June to less than 750 in February.

Listening habits based on the time of the day:

Next up: I have gone a bit deeper and tried to decipher whether I have listened to more songs at day or at night. For that, I have classified Day and Night in the following way:

  • If the time is after 5 am and before 5 pm then it is day
  • If it is after 5 pm and before 5 am then it is night

Based on this assumption, this is how I have listened to songs with respect to each month:

Nothing is very obvious from this plot but few points:

  • For most of the months, the proportion of songs that were Scrobbled in day is higher.
  • The proportion of songs for night is higher for June and July. It makes sense, because during the lockdown in 2020, I used to stay awake for long!
  • For February to March, the proportion for night is lower than other months. During this time, I started going to the University, then had Covid and the got the Job. So, it is also expected! I try to follow a routine, you see.

From the box plot for the day and night it can be seen that median number of total songs listened during that time is higher for the day than for the night. There have been also many outliers for night compared to that for day. It suggests, there have been many nights when I have listened to more songs than I usually do at nights. Don’t we all have those nights where we immerse ourselves with the melody of our favourite artists, and lost the track of time?

GitHub link for the file: