In my last post, https://lakedatainsights.com/2018/10/14/why-data-science/, I described why I got interested in data science in 2014 and how I created a learning plan to get started in the field. As I started looking around for learning resources, I came across a very valuable source of learning – Massively Open Online Courses, or MOOCs (https://en.wikipedia.org/wiki/Massive_open_online_course).
MOOCs can be a great way to get world class education for free. Sounds pretty amazing. There’s one challenge, however, and that has to do with follow through. A quick search of “MOOC completion” rates indicate that less than 10% of people that sign up for a MOOC class actually complete the class. Well, I’m one of those 10% for two classes that I signed up for where I learned a lot. There have been a couple other classes where I sit comfortably in the other 90%!
The first class I took was an Intro to Data Science course taught by the University of Washington through the Coursera platform. I don’t see this specific class on their site now, but there is a Data Science at Scale Specialization available (https://www.coursera.org/specializations/data-science). The course I took was a 9 week course that was a survey of topics in data science – map reduce, visualization, data base technologies, machine learning, etc. There were programming assignments for most weeks along with on-line lectures.
This UW class was a good class overall. Looking back through the content four years later (I took the class in 2014), I realize that it covered a ton of ground in 9 weeks. As one might expect, there wasn’t opportunity to dive too deeply into any given topic. One challenge that I recall was the variety of technologies covered in the class. A significant amount of time was spent getting tools installed and getting familiar enough with them to do the problems.
The second class I took was called The Analytics Edge and was offered by MITx. It was offered in 2015 on the EDX platform and was a very highly reviewed class at the time. While still an intro class, it wasn’t as broad as the UW class and the technology ramp up week to week wasn’t as intense. A primary focus was machine learning, both supervised and unsupervised. But it also had a week on visualization and two weeks on optimization.
I can see why the MITx class was highly reviewed as I learned a lot in it. There were a ton of good examples in the class and the consistent use of R throughout the class enabled me to build that skill week over week. For that class, I decided to pay $100 to get a certificate after the class to demonstrate that I did all of the work satisfactorily.
There have been several certificate programs that have shown up on the MOOC platforms in the last couple of years. Some of them are aligned with a company and its technology; Microsoft has a number of classes on the EDX platform, for example. Others are aligned with universities, such as the aforementioned University of Washington program. There are even graduate programs available on the MOOC platforms.
Can you become proficient in data science using MOOCs as the primary education tool? It’s definitely possible, in my opinion, especially when coupled with project work on the job or as a hobby. But it takes a lot of discipline. And it takes some good planning since you have to come up with the curriculum on your own. Based on my limited exposure, I would lean towards the classes offered by universities over those offered by technology companies.
But you don’t have to take my word for it. Here are three other data points on the use of MOOCs for learning data science.
- In https://hbr.org/2018/10/the-chairman-of-nokia-on-ensuring-every-employee-has-a-basic-understanding-of-machine-learning-including-him, the chairman of Nokia describes how he set an example for his company by personally gaining machine learning knowledge. He ended up taking six courses in three months, doing a project for each. I can’t quite rationalize how he managed that rate of course taking when it took me nearly three months for each course I took… other than to observe that he’s the chairman of Nokia and I’m not!
- In https://simplystatistics.org/2018/10/08/the-economic-consequences-of-moocs/, Jeff Leek reviews a paper that studied the economic impact of taking data science MOOCs. The payback for those that completed courses is impressive.
- In https://datamovesme.com/2018/08/01/favorite-moocs-for-data-scientists/, Kristin Kehrer reports on a survey she did on LinkedIn regarding favorite MOOCs for data science. Check out the post for the top responses for a several technologies.
I ultimately chose a different route, but MOOCs were a great way to get started and a powerful continual learning tool.