Write a Review: The Analytics Edge

Tell us about your experience

 
 

Click stars to rate

150 characters minimum

Tell us about yourself

Progress in the course

Prior knowledge of the subject

First Name

Last Name

See what other students say

Student profile image
Student profile image

Student

6/10 starsCompleted
5 years, 9 months ago
I agree with all the other reviews regarding the course material: the lectures are great and cover fantastic concrete applications of data analytics in business. The Kaggle competition was another brilliant ... I agree with all the other reviews regarding the course material: the lectures are great and cover fantastic concrete applications of data analytics in business. The Kaggle competition was another brilliant idea perfectly executed. However, despite these strengths, the course experience is severly damaged by the length of the weekly problems. I sometimes have the feeling I am the only one thinking this way, but I'll share my opinion anyway. The problems contain an insane amount of trivial questions that usually require nothing more than copying/pasting a formula available in the source code provided with the lectures. Don't get me wrong: I am fine investing a significant amount of time in a MOOC. My point here is that the time spent on the problems brought no added value. The questions offer no challenge: the datasets are always perfectly clean and formatted. The questions are like "Type summary(dataset). What value can you read for field ... ?" The way the problems were framed gave the impression that data science is a clean, linear, deterministic process that offers no challenge, no suprise and no headache. The problems were not leveraged to introduce basic R concepts such as FOR loops or IF conditions. They were not used either to show how data sets often come sparse, heterogeneous, dirty, hard to decipher not to say "hostile". Key concepts such as overfitting and cross validation were barely mentionned, while they should have been exposed through specific dedicated problems. With so much of real data science tasks and concepts left aside, many students felt helpless when the Kaggle competition started. Many had no clue what to do with missing values, what to do with "Yes/No" values: convert to 0/1 ? -1/1 ? Should we scale ? What should we do with obviously wrong data (Year of brith in the future) ? There are more than 100 features available. Is that too many ? How do I know if that's too many ? I want to try a model with fewer features. How do I select the "best" ones ? My model performs great on the training set but not on the test set. What happens ? How do I fix that ? Etc. Unsurprisingly, so many hours spent answering trivial closed questions and no methodolical open questions produced very little actionable skills and experience when the time for Kaggle came. It may sound hard, but I am optimistic: the problems can be rebalanced pretty easily for the next iteration of the course, should the staff agree with the diagnostic. As so many others have said: the rest of the material is fantastic.
Was this review helpful? Yes15
 Flag
Amit Gairola profile image
Amit Gairola profile image

Amit Gairola

8/10 starsCompleted
2 years ago
The course was excellent in terms of explaining the concepts. The large number of datasets and examples was good to learn. The homeworks and assignments were quite elaborate. The section on Optimization was ... The course was excellent in terms of explaining the concepts. The large number of datasets and examples was good to learn. The homeworks and assignments were quite elaborate. The section on Optimization was really good. However, there are some important concepts such as Time Series, Naive Bayes, Neural Networks etc which are lacking in the course.
Was this review helpful? Yes0
 Flag