Learner Reviews & Feedback for Introduction to Data Science in Python by University of Michigan

The lectures are insufficient, the homework instructions are incorrect, the autograder is terrible.

the teaching materials provided by christopher is too brief. can be improve by more elaboration.

pathetic course. Not for anyone who wants to learn python. Its for them who already know python

Very tough question asked without proper teaching. Course should be made learning oriented.

very hard assignment and confusing that i have to withdraw from doing :(

Instructor is not interested in teaching but just reading out the script

This course is not suited for beginners, who wanted to learn python.

Horrible. Explains nothingJohns Hopkins data science is FAR superior

i have been unable to follow the course and i want to exit

Poorly organised. No helpful resource. Really sad course.

Vedio is useless, I only need to go through the code.

the teaching pace is very fast and not so clear.

Gap between lecture and assignment is too high.

Disconnected, too broad, could have been great.

Worst class ever! The instructor sucks!

Not exactly benefited.Too fast

autograding system is terrible

assigments are not clear

The speed is too fast

not good course

It's boring.

outdated

IN SHORT

This was a great course and I learnt a lot! Topics covered include a quick reminder on intermediate python and lots on pandas and some numpy. The weeks 3 and 4 assignments are quite challenging so expect to spend considerably more time than indicated on the course site if you're not experienced with python and pandas. This course is not for coding newbies.

IN DETAIL

I am proficient in R for data analysis and had dabbled with python before although had no experience with pandas. I was committed to learn the course material and to spend a substantial amount of time doing so. The speed of lectures is fast. I paused often to take notes and to try out the provided notebooks, and I returned to some of the videos when working on the assignments. I found the course assignments good and challenging. The lectures give a good tour of different functions and approaches you may want to use in the assignments, but there isn't much handholding with the assignments and you'll most likely spend quite of bit of time looking things up online in pandas docs and stackoverflow. If you're used to that and generally troubleshooting code, you'll probably be just fine. I spent much more time on the assignments than what is estimated on the site: ~5h for week 2 (vs 1.5h indicated), ~1 day for week 3 (vs 2h), and 2.5 days for week 4 (vs 4h).

Week 1 gives a refresher on how to write functions, list comprehensions, and lambdas in python. If you're familiar with writing loops and functions in other languages, with this material you will get to writing them in python quickly if you invest a bit of time and effort. If you're not yet at the level of confidently writing functions, loops and vectorized alternatives in python or another language, I'd recommend starting with a different, more basic course because the learning curve with this one might be too steep.

Week 2 gives the ins and outs of pandas including creating and querying pandas series and data frames.

Week 3 covers merging data frames, grouping (groupby) with aggregation (agg), applying functions rowwise (apply), and pivoting data (pivot_table) etc. It also gives a whirlwind tour of date/time manipulation using pandas. numpy is also included.

Week 4 has some lectures on distributions and more on numpy. The week consists mainly of the main project assignment where 50% of points are given on data cleaning and munging (contents of weeks 1-3) and the other 50% of points are on modelling and hypothesis testing. It's quite a proper project in the sense that you're given a number of non-clean data files scraped from different places and a hypothesis to test. There are some additional instructions on what format of cleaned data to produce from the different files and what type of test to perform, but for the rest you're on your own.

This course was excellent. This course deviates from many garbage MOOC who only work with quiz and can not provide a real python coding challenge experience. Assignements are really tough. But my sense of progress is real.(I have struggled to identify such feel in many pytyhon MOOC). Jupyter base for everythjng is a fantatsic format(it even allows coding mobility betwwen my station at work and my home station through the coding on jupyter in the cloud) . My feedback nevertheless will point to some aspect in my experience and where I think you can improve.

Succeeding the assignement does not mean that we identified the most elegant way to apply all the knowledge of the course(lambdas,list comprehension, grouping..., apply) in our coding. Breaking that barrier is not easy for me unless we are forced at it and so my looping mind is often applied in assignments. A real correction with the answer need to be provided(this is what the real classroom would do, we managed to get to the answer but we could still learn more with an assisted correction just like what the real classroom would do.I understand that you are worry that the model will end up as copy paste on a webpage and will kill your value. You could maybe consider this add_on for paid customers only and only provide it in picture way which can only be paper print and not so easily converted to webpage format.Or you need to find an alegant way to randomize the assignment coding test at each coursera session, which in that case would not bring any forgery issue and you could provide the correction at the end of the course(or after each assigment completed).

Videos are a bit too fast on concepts sometimes.

You could split the assigment in two formats: format where simple principle of the course are first resolved on jupyter notebook (just like the videos case but with more exercices) and complex dataframe case as second assignment .(but please reduce the amounts of case to only 1 or 2, not 3)

You could reduce dataframe case.(I've spend easily 40 hours on assigment here, assigment time is too heavy from my workload as a full time scientist. This needs some carefull tuning.

Overall Great Job

Very nice Course, You will Learn about how to effectively use Pandas Library for Python and how to treat DataFrames in that ambient, there are nice functions and methods for parsing. The Course is very fast pace, I only have time on the weekends (some of those), so I had to switch dates two times. Also, some materials are very fast, so If you are new in Python, got to be sure if you have mastered prior concepts of the course (Week 2 depends on week 1 and so on ...). A large part of the course involves your own research in Python Docs and StackOverFlow page. As I am an R user, some things are intuitive (and maybe more easier for me to do in R), several of the things in comparison I thought: "Wow, but this is so much easier on R", but at other times I saw the power of Python for parsing tasks or webdata that in the R require too much memory or are more complicated to obtain. R have the problem to treat everything like an object and guided by vectors, but at some parts that makes understand coding details more deeply, at least for me. I still think that the documentation of R the best there is for my purposes, so I will stick with that, but it's great to know how to develop some things in Python, mainly because of my goal of getting some applications to end users. Finally, not much related to the course, but maybe with the change in the platform of the Coursera, the forums seemed a little more confusing and a little more slower than former courses that I took, I think it might have to do with the same course running in parallel on different dates.