The world’s leading publication for data science, AI, and ML professionals.

How I’m Overcoming My Fear of Math to Learn Data Science

Fear is no reason to not pursue a passion, especially if that fear is all in your head.

Office Hours

As someone who failed math in high school, I feel particularly qualified to write this article.

I didn’t know much about the Data Science world until I stumbled onto Medium and discovered the Towards Data Science publication. From the first article I read, I was hooked.

Data science incorporates everything I love about tech with science, business, medicine, engineering, and just about any other field you can think of. As someone trying to get into the environmental consulting world, I knew that data science could be the arrow in my quiver that seals the deal.

I knew that I had to at least try my hand at data science.

However, there was one thing that dampened my excitement: the requirement of a deep understanding of mathematics and/or the ability to teach oneself advanced level calculus, linear algebra, and statistics.

As I mentioned above, I failed high school math. I was not starting off on the right foot.

I began getting overwhelmed as I browsed through articles that laid out how much math you needed to become a data scientist. Generally, those articles listed pre-calculus, calculus, multivariable calculus, trigonometry, linear algebra, differential equations, and statistics.

A weird fact about me is that I’m not inherently bad at math. I’m just bad at math that isn’t practical or applied. Math that occurs in statistics, physics, chemistry, or calculus, I have no problem working with. To be honest, I’m actually really good at it. But when it comes to more abstract applications, such as number theory and differential equations, I’m screwed.

Because of this, I was dreading a good portion of the mathematics I was supposed to learn to become this almighty data scientist.

That is until I came across this article:

You don’t need to know much math for data science

The discovery.

When I first read the title, I disregarded it as blasphemous. At the time, I thought anyone who claimed that you didn’t need to know much math for data science was just looking for those clickbait views.

How was it possible that this article could be preaching the complete opposite of all these well-received articles that I had been reading? Was everything that I had been reading misinformation?

As I continued to read the article, I noticed that the author had made a distinction that hadn’t been made in any other article I had read on the subject: there’s a difference between theoretical data science and practical data science.

In short, the article explains how theoretical data science (practiced usually by academics) is quite different than practical data science (which is usually practiced by industry professionals). Data science that is practiced in academia is, generally, much more mathematically intense than data science practiced in the industry.

While the author admits that the level of math you need will vary depending on your seniority level as a data scientist (junior versus senior), one of the more telling factors is whether you will be working in academia or industry. Of course, depending on the industry you’ll be breaking into, you may need more math than in other industries. However, that level of math is still likely much less than if you were to work in academia.

The author further goes on to explain how foundational data science skills, including data manipulation, data visualization, and exploratory data analysis, don’t actually require much math. The process of collecting and cleaning data and producing scatterplots or histograms hardly requires advanced level mathematics, and can even be completed (in the author’s opinion) using the math skills you learned in high school.

When it comes to machine learning, one of the cornerstones of data science, the author explains how there is a misconception that excellent Machine Learning practitioners must have an advanced understanding of math. This misconception is untrue, as the author describes how several machine learning practitioners that he knows personally have no advanced math training, yet still work at major companies such as Apple and Bank of America. What those practitioners lack in theoretical knowledge, they make up for in their ability to apply mathematical techniques.

Furthermore, like with other data science applications, there is also a difference in the setting in which you intend on practicing machine learning. Working in academia or industry will influence the amount of math you need to know to create successful machine learning models.

The author uses the example of reading an academic machine learning paper versus looking at the results of a machine learning model created by an industry professional. The machine learning paper will include tons of advanced level math, whereas the practical model will likely only involve basic statistics, linear algebra, and calculus – otherwise known as the level of math you may learn at the undergraduate level in university.

By boldly stating that data science and machine learning practices don’t require years of intense professional mathematical study, the author flung the doors open for up-and-coming data science practitioners like myself who otherwise may have looked the other way due to a fear of math.

So what math skills do you actually need to start learning and becoming a competent data scientist? The author recommended five simple areas of Mathematics in which to focus your study:

  1. Basic charts and graphs.
  2. Functions.
  3. Basic algebra.
  4. Basic statistics.
  5. Basic math notation.

For aspiring junior data scientists, this list of mathematical competencies is more than enough to solve any problems you’re going to come across. I’ve read many times how experienced data scientists value going back to the basics and using the simplest solution to solve a problem. Those simple solutions likely involve using an Excel spreadsheet and some simple algorithms to yield a result. In other words, not every data science problem requires differential equations, complex machine learning models, and artificial intelligence.

How this is changing the way I learn data science.

Back in December 2020, I published an article highlighting my new year’s resolution of learning data science in 2021. In that article, I highlighted the learning curriculum I had created for myself. At the time, my understanding of data science had been driven by articles written by top-level senior data scientists who included laundry lists of mathematical competencies that everyone should know if they want to be a data scientist.

Now, after reading that article, I’m more inclined to restructure my curriculum to focus on more practical skills, particularly descriptive and inferential statistics, basic calculus, and fundamental data science competencies. In short, the mathematical portion of my curriculum will look a little more like this:

Mathematics

  • Finite Mathematics
  • Descriptive and Inferential Statistics
  • Calculus
  • Algorithms/Techniques: linear regression, logistic regression, support vector machines, decision trees, neural networks, regularization, and principal component analysis

Furthermore, I’ll be focusing my attention on going straight into learning different data science and machine learning algorithms. While it’s important to know why something works, I’ve also learned in my past math education that it’s okay to just understand how something works. So, instead of getting caught up in months of mathematical study, I’ll focus on how to use a particular algorithm, when to use it, and the results it will yield. This practical way of learning math will reduce the noise, and simplify my understanding of the most important details. At some point, it will be time to learn the why, but until then, I can get by just knowing how to properly apply the algorithm.

Basically, I’m narrowing my focus to the most important concepts. I want to focus on cleaning and manipulating data, visualizing data, and performing exploratory data analyses. By focusing on the four foundational data science skills, I will be competitive in the industry space, not as a data scientist in title, but as someone who can apply data science principles to solve problems.


Final thoughts.

As someone who’s been scared away from professions before due to a lack of confidence in mathematical abilities, I feel like it’s important to look at any discipline from every angle.

In the case of data science, everyone is so obsessed with working under the "data scientist" title that they put themselves through a world of hurt trying to learn complex mathematics. In reality, most people that are afraid of math but want to work as practitioners should focus more on learning the practical side of the trade, and then applying that to roles outside of the "data scientist" realm. It’s been proven time after time that individuals who learn technical skills will get pay raises at work, just because they can practically apply what they’ve learned in a way to benefit their company.

Therefore, maybe the target for those of us who are afraid of math is to become practitioners of data science, people who can clean, manipulate, visualize, and analyze data in the context of our given profession, and do so without having to undergo months of training in a subject we can’t often wrap our heads around. By knowing the foundational aspects of data science and knowing them well, we can make an impact – regardless of whether we can do differential equations or not.


Related Articles