Learner Reviews & Feedback for Applied Social Network Analysis in Python by University of Michigan

Going into this course, I was really disappointed that I had to take this course for a Data Science Specialization because at a skin-deep level it seemed very irrelevant, and frankly I was at that state of mind until week 4 of this course.

There are several reasons why I'm rating this course 2 out of 5 stars:

1) The content of the first three weeks were just informational and should have been covered in one or two weeks.

2) Homework assignments were not challenging at all. 90% of the questions were one-liners and required simply calling the methods of networkx that was discussed. This course would benefit by homework assignments that had 1-2 problems that required us to solve real-life problems from scratch, rather than ONLY calling networkx methods.

3) There was no discussion on how to get network data. We were just given all this magical data about how relationship scores between employees and future connections between employees... How am I supposed to get that in real life?? Some problems asking us to make a network would've been valuable.

4) More time should have been spent on prediction and other advanced topics, at least another week to bring the "Applied" into "Applied Social Network Analysis."

5) I really enjoyed the professor's teaching style. He explained concepts well and had great examples during lectures.

If you are following Dr. Severance's Python course series, expecting a similar experience, be ready for disappointment. Where Dr. Chuck's series is well-taught in-depth information with great examples and explanations, this course series is basically a summary and the instruction to go out and learn everything yourself from the web. That's an abysmal way to teach, because for beginners to data science programming, there is no basis to determine which is bad or good information you pick up from a blog or Youtube.

This class is a perfect example of just that, but worse, because there's an implicit trust in a professor from a university the caliber of Michigan.

The lessons are old, and they won't update them. The forums are full of bug reports and information about code that works when tested outside the class environment, because the class is taught on networkx version 1.11, which was released in 2016. I'm writing this in April of 2020, and networkx 2 was released almost 4 years ago, and this class is teaching techniques that no longer work because the code is deprecated.

In every course in this series, I wasted long hours trying to debug what I was doing wrong in my assignments, only to find out that my code was perfect - today, and the grader was testing my work with functions that are no longer valid.

I know reviews can be biased, especially negative reviews, so please read the forums before enrolling in this course and wasting your money. The quality of education from this series is very poor, and you are better suited to learning from Youtube (sadly) if you want actual information you can use in your work.

Unless you are looking for an education so that you can go back in time and find a job in 2015, this course is a waste of time and money.

This course is a good introduction to graph theory. Its contents are interesting and the lecturer did a great job explaining them. So, what is the problem? The problem is that the course is not called "Applied Graph Analysis in Python" but "Applied Social Network Analysis in Python". This incongruity in the title of the course (intentional or not) will generate erroneous expectations in the students, especially if we consider that they have to take the course to finish the specialization. Regarding the assignments, they are divided into two groups: trivial tasks that are solved with a single line of code extracted from the NetworkX manual and more complex tasks related to Machine Learning that do not involve putting into practice the concepts of this course but those of the third course of the specialization. I regret being so tough, but my impression on this course is that it is filler content designed just to have a five course specialization instead of four.

The specialization for Applied Data science started strong, with engaging exercises, good instruction, and good recommendations for additional reading and resources. As the specialization continued, the courses seemed to get "lazy", and the course topics became more abstract and less applied.

After going through this specialization, I would not recommend this to someone if I could find a better program through edX or another coursera offering.

The lectures are good. However, the assignments are poor: very simple exercises with toy examples, but far away from real applications. Moreover, I spent most of the time (particularly in the last assignment) trying to deal with the autograder.

For a coding heavy course, why doesn't the instructor just upload the code used in slides as a Jupyter Notebook? This would save A LOT OF TIME and frustration. Right now, I have to pause the video to copy the code AND write my own notes and it wastes so much time. Not to mention, you can easily be prone to writing wrong syntax when you're trying to keep up so fast, and then you run the code chunk and it doesn't work and you have to go back to that point in the video. It's a simple staple that I would have expected in a UMich course. Also, they don't show how to create networks from pre-existing data, which is how you will usually work in the real-world

The assignment auto grader was horribly designed.

This is hands down the best taught course in the speciality. The instructor explains concepts in the videos clearly and the assignment questions are structured and interesting. Do note that the assignment in week 4 does pull together the whole specialisation in a real world problem, so if you aren't taking the whole speciality you will need a knowledge of Pandas and SKLearn. Personally I thought it was pitched at just the right level because the ML work is just enough to have to go through the process, without any complicated feature optimisation.

Only wish the other courses worked as well as this one.

It was an easy introductory course that is well structured and well explained. Took me roughly a weekend and I thoroughly enjoyed it. Hope the professor follows up with more advanced material.

Excellent tour through the basic terminology and key metrics of Graphs, with a lot of help from the networkX library that simplifies many, otherwise tough, tasks, calculations and processes.

Really enjoyed the mathematical component of this course. It was fun to see how you could connect the graph theoretical components to the machine learning concepts from earlier courses.

This is by far my favorite Coursera course - well organized contents and intuitive example!

Great course, maybe even the best on this great specialization!

Maybe I was getting used to the grader, and getting better at Python, but I found this the least frustrating of the specialisation, with a lot less time wasted on grader issues and the like.

Its a good general introduction to the theory, with some good exercises which combine network analysis and machine learning.

Like all of these classes, it is getting a little long in the tooth (2016, so a good four years old now, in a very fastmoving field), and it would be nice if it was updated a little, but given that its basic stuff, it is still very relevant.

It would also have been nice to have had a challenging, bring it all together, project at the end of this, but I guess its my job to find one now!

Well made intro to networkx library and some basic graph concepts.

It seems that course material has been done 3 years ago, and has not been touched since. The age is showing bit. For example autograder has small issues which could be fixed to remove lot of frustration and extra work from the studet

For example sometimes when returning pandas Series, the autograder requires that it doesn't have a name, but the autograders feedback doesn't give any indication about this. Minus one star because of these. All the courses in Applied Social Network Analysis in Python by University of Michican has this same issue.

I think that assignments 1-3 were too basic; often, you only had to return a simple function which outputs a specific network metric. Assignment 4 was a lot better, as it comprised the necessity to apply knowledge from previous courses. The instructor did a good job explaining the different concepts!

it is good but we are looking for more real practices

They need to change the 4th assignment is almost impossible to run on jupyter

Week 2's quizzes are highly irrelevant to the content taught.

Well, the actual score is more like 4.5 stars since I still rate the Machine Learning course by Andrew Ng as the best course I've taken on Coursera. Anyway, this course in my opinion is the best course in the specialisation and I'm glad I stuck around for it.

Pros: the instructor has a very good delivery, and explains concepts in sufficient depth gradually, I really like the way he explained how some measures are calculate using step-by-step examples showing which nodes/edges are being used. Compared to Professor Brooks who either gives very superficial lectures on what matplotlib can do (line graphs) to suddenly going into the technical details of the different matplotlib layers

The assignments are the most reasonable I've seen in this specialisation, tying relatively well with the course lectures (though Week 3 assignment's might be tad too simplistic).

Con: Outdated autograder as usual, however this was probably the mildest case compared to Course 1 (Intro to Data Science) where so many things were different between the old and new pandas version.

Here's my personal overall ranking of each course in this specialisation from best to worse:

1. Social Network Analysis

2. Applied Machine Learning

3. Applied Text Mining

4. Intro to Data Science

5. Plotting

So, I passed all modules in the whole specialization and received the certificate. This is by far the best course, and the reason for this is the instructor. Daniel Romero is great at explaining the concepts, expresses himself clearly and uses lots of examples which help immensely. The programming assignments are actually fun to solve - the instructions are clear and well-formulated. I know what is expected and can focus on doing data science. For the first time I didn't have to spend hours reading the Discussion Group posts in despair, in order to figure out how to pass the assignments (tricks, hacks, etc). This can't be said about assignments in other modules. I think the assignments were not too easy - to me the difficulty was just right. It's an introductory course to this matter and the worst you can do is daunt learners with unrealistic assignments (as in Week 4 of Text Mining). I think my appreciation for this course is intensified by the irritation with other courses. But at any rate, great job Daniel.

Daniel Romero is easily the best instructor in this specialization, beating all his more senior colleagues in communicating knowledge to learners. I especially liked his methodical approach to telling the whole story: big picture intuition, formal definitions, Python examples and practical applications.

The course sets you up with a solid enough baseline in networks to be able to continue learning on your own. It is a big plus that the final assignment is really pulling together your knowledge from most courses in this specialization (there is no visualization required in the final assignment).

My only complaint is that the course seems like somewhat of an afterthought to bring the course count to 5 in the specialization. The content in week 4 is particularly short. I would rate this course 4 stars on its own, but given how it lines up in the specialization it is an easy 5-star.

nice course with good content in quizzes and assignments. The last assignment was great and very practical (well framed question, which uses ML algorithms to predict node attributes and linkage using various network measures as features). Overall a pretty much useful course using graph theory and a practical course. Most of the assignments are concentrated on how real world problems could be?

The centrality measures are explained beautifully. The module four is pretty much the heart of this course.

I have really enjoyed learning this course. All the concepts are explained with proper examples. This course not only provides theoretical knowledge about network analysis but also explains the use of each topic in real networks. The assignments were really helpful to get hands on experience of all the topics covered. The most interesting part of this course was the last assignment It was fun experimenting with different models and analyzing the performance.

This is the last course of the Applied Data Sci in Python certificate. It effectively ties together all the introduced concepts from the previous courses (except Natural Language Processing). Daniel Romero was an extremely effective lecturer and many of the concepts and know-how were introduced, taught, and assessed appropriately. I'm also impressed that I was able to learn a new python library I (or my coworkers) had not heard of before.