The Analytics Edge

As a software engineer interested in ML techniques and algorithms, I did not enjoy this course. I had previously completed the popular coursera ML course by Andrew Ng which I enjoyed, and I was hoping in this course to both get familiar with R as well as get my hands dirty with real-world scenarios.

Unfortunately, this course was a painful approach to learning R and analytics, and I can't help but feel that it could've been done much better. My complaints:

- R: Not much time spent familiarizing with the basics of R, and instead being thrown into the world of R analytic libraries, each library having its own arbitrary syntax and function parameters. I ended up really disliking R.

- A very poor way of teaching how to play with data. The student isn't challenged to "figure out what to do next with this data" and is instead spoon-fed the next step, having only to reproduce the correct library/function/syntax from memory. Learning by repetition, not by deeper understanding.

- The step-by-step assignments are also designed such that you have enter the results of each step one at a time, in order to verify that you're getting the right results, scoring micro-points along the way. Time-consuming and cannot be sped up. The weekly lecture videos can be watched in 1-2 hours, but the assignments/recitations take 8+ hours.

- A very poor presentation of the theory behind the techniques. Slides are presented and mostly simply read off of. I'm sorry but this course needs to find better teachers. A couple of the recitation TAs were pretty good though and I appreciated them more.

On the positive side, what I got out of the course:

- Some hands-on experience with real-world data analysis. Reality-check that most applications of data analysis don't involve chinese boardgames and self-driving cars.

- Some familiarity with R, contrasting it with Octave/Matlab, and also realizing that there are libraries for everything.

MIT’s The Analytics Edge is an edX course focused on using statistical tools to gain insight about data and make predictions. The majority of the course teaches analytic methods using the R programming language, but the final 2 weeks deal with solving optimization problems using spreadsheet software (LibreOffice or MS Excel). The course runs 11 weeks and covers R basics, linear regression, logistic regression, decision trees, text analytics, clustering, visualizations and both linear and integer optimizations.

The Analytics Edge is a meaty course. It has a lot of content each week and it’s not easy to breeze through things like it is with many other MOOCs. There are graded quizzes after each video lecture and each week of new material has 4 fairly lengthy case studies to complete. One week is devoted to an analytics competition while the final week is reserved for a 4 part final exam. Some students on the forums claimed they were spending 10 to 15 hours a week on this course. Coming into the course with basic knowledge of statistics and R helps a lot. It should be noted, however, that this course is not too math intensive. It doesn't spend a lot of time talking about formulas or nitty-gritty mathematical details; it mostly teaches you how to apply statistical functions and methods and interpret the results.

I didn't take this course for credit or certificate because I already have a MS in EE and an MBA, and I was taking other classes simultaneously. My goal was to "skim" the content for expansion in the future. However, the content and exercises were so well organized (most step-by-step) and relevant to real-world problems that I ended up spending lots of time understanding the material and writing lots of R code that I archived for future reference. The course wasn't terribly difficult, but there was a lot of material. I skipped the optional lessons until I completed the course, but now I am going back and doing the optional lessons. I did the Kaggle competition and finished in the middle of the pack (in the lower 0.6xx accuracy). To get in the "leader" category (two in the 0.9xx accuracy) will require a lot of work and knowledge of R beyond what is covered by the course. If you want to be among the leaders, be prepared to do a lot of Web searches and searching R documentation. I learned a lot, but I'm still very humble and respectful of the experts.

This is one of the best courses on the topic of Analytics/Data Science/Machine Learning that is out there. It is also one of the best MOOCs that I took. Its approach to the mentioned field is the "black box" approach. In other words, the course will not teach you how to build an ML model from scratch nor what is exactly being done in the background. However, it will teach you how to apply the already existing algorithms to analyze data and interpret your results. And, it will teach you that quite well.

For the majority of time, you will be dealing with Data Science/Machine Learning methods, but in the last two weeks you will learn about Mathematical Optimization as well. Each weekly assignment will consist of 2-3 cases in which you will apply the learned methods and derive results. And for me, here comes the best part of the course. For each case in the assignment, you are not just given a data-set and told to calculate something. Namely, you will not experience the situation: "Where do I even start ?" The case questions lead you progressively from understanding the data-set and answering basic questions, to answering advanced questions which shape the consequent, real-word decision (if you would be the one making it, of course).

Even though the course is great, it is wrong to think that you will become an expert in DS/ML just by finishing the course. However, I personally find it as a great source when entering the field, or even for enhancing your knowledge/skills. It is probably at an intermediate level in terms of difficulty. The work is being done in R for the DS/ML part, while the Math Optimization part is done in spreadsheet software (Excel, Libre Office, Open Office). The only downside that could be mentioned is that all the R exercise videos have been made in R Console and not the R Studio (and the video quality is not HD), but you can easily follow them using R Studio, what you should definitely do :) At the end, I can highly recommend the course to all those who want to learn more about the amazing field of Analytics/Data Science/Machine Learning.

I took this class in 2016.

I joined this class randomly one day, and it was literally the first time I became aware of data science as a field. The class also hit me at a time in my life that I was ready for some big changes.

Over the next several months, the class helped me to see the amazing potential, necessity, and inevitability of analytics. As someone with only high school math (from so long ago) and no programming knowledge, the class was very challenging but the topics were engaging. Some weeks, I put over 20 hours. "Argh, you said I needed only high school math!" I wanted to give up so many times, but the satisfaction I got from *finally* figuring it out fueled my momentum forward. I even managed to finish the course on time, which I wasn't expecting to do. It pushed my mental limits and reshaped my self-esteem with STEM subjects. I'm so glad I stuck through it.

The reason I give only three stars is that--- well this scoring thing is purely personal, and personally, I don't favor it too much because I like more technical stuff. As it is offered by the Business school of MIT, for me, taking it might be a wrong choice. But I wouldn't be so certain about my tendency before taking it, so it offers a diagnostic value to me as well.

I finished this course around June, 2017. This course is a good choice for people who want to be more on the business side rather than the technical side.
* Technically it is nothing more than applying a bunch of R packages on data.
* But in terms of application, it offers a good variety of context of problems, they are real, authentic. And it comes with mountains of exercise.

For business people, I believe taking this course can be like empowering yourself with a new toolbox.

The strength of this course is its broad selection of USE CASES compared to other data analysis MOOCs, some of which have a more technical focus. If you are interested in getting a taster of what analytics can do, willing to play with a bit of code but do not wish to spend much time cleaning data, this offering from MIT Sloane is worth investing in.

As a beginner to R, with little coding background I found the pace fine. I did need to spend a bit more than the suggested 10-15 hours / week to get though all the material, although you can be selective and only finish parts of the syllabus to complete the course. Some of the homework activities are not as well-constructed as others, this was probably to take into account the diverse student level. Community discussion was vibrant.