It needs to be updated. Its probably one of the most in demand skills in the field and this has a weeks worth of content 1 section 25 minutes of video 5 questions. Its just not as good as any of the other courses. 100% needs to be revamped.

By far the laziest course set up in the track. It's an interesting topic, but without independent study I would have learned almost nothing due to the lack of any "practicals" in this "Practical Machine Learning". A really disappointing course that fails to be worth more than just a couple hours of youtube.

In general, great course. But because of the strong interest in ML, I am going to attempt a detailed review.

PROS:

This course truly de-mystifies "Machine Learning". After completing the course, you will be able to programmatically use 100s of ML algorithms that have been created by others over the years. You will be able to use the Caret package in R to simplify your application, simplify pre-processing, perform automatic cross-validation/model tuning and generate various statistics about the model used by your ML algorithm. You will be able to easily estimate out-of-sample accuracy to determine if your model has any hope of working well, picking one classifier over the other, or using several classifiers to estimate outcome. You will learn how some of the heavily used algorithms in the industry work behind the scenes, and where to go to learn more about these. Several learning databases are introduced. If you tinker with them, you will be amazed at how easy R and Caret make it to apply ML algorithms. You will understand how chatbots, recommender systems, spam filters, "prediction" systems and the like work.

WHAT THIS COURSE DOES NOT COVER:

It does not cover how to write your own ML algorithms. That requires working knowledge of optimization algorithms, advanced math and probably lots of other resources.

WHO SHOULD TAKE THIS COURSE?

Only those prepared to work hard, dig in, and persevere through a lot of (sometimes difficult) material will benefit from this course. If you're not confident about your statistics concepts, not comfortable with R and databases, not comfortable with googling for parameters and techniques not directly discussed in class slides, then you will have trouble. Passing the quizes require you to refer to material from prior weeks, read online documents and look for similar solutions at stackexchange etc.

TIP FOR MENTORS:

For every week of the course, create a pinned post which says "Tips/Errata for Quiz #n". You've collected sufficient feedback from students now and know what the common issues are. Don't make them search through 100s of discussions to figure out solutions to well-known/common problems.

More practice Items are needed

Worst lectures! Worst teaching! I leanrt most of the topics on statquest. Very very very highlevel teaching, very little effort put in by Bcaffo and Rdpeng on this! So many issues in the quizzes. Wasted hours on puzzling out what is to be done! Have a look at the complaints in the course era discussion board. Issues since 3 years are not corrected. The course needs an update. But no m*****F**** is listening. Solutions to quizze are wrong! I have had it with coursera and their useles peer correction. You don't even know if what you are doing is right! Worst FEEDBACK ever!

The worst course of the specialisation so far. The quizzes are full of typos, not clear at all, and the videos teach nothing, always refering to elements of statistical learning book. Now that I have completed the course, I do know a bunch of algorithm names involved in machine learning, but I certainly do not understand what they do and when using them.

The lectures and quizzes are based on old versions of R and R packages. This course needs a serious update, as some packages work differently, test answers have changed (but not been updated) and coding along with the videos results in different results. Going to the forum you can see that this has been an issue for a few years now.

Unsatisfactory and poor course in this specialisation. There are many important parts which are explained inaccurately. In many cases, the lecturer jumps from important points, or assumes students have detailed knowledge about the topic. You can find ambiguity in weekly questions. Very unsatisfied!

That's a pretty rushed course. I think you really should reformulate it and discuss its content with a deeper way.

Lectures and course material is insufficient to get the right amount of knowledge to be able to do the tests and the course project

I'm a big fan of the John's Hopkins Data Science series on Coursera; however, they definitely "phoned it in" on this particular course. No practical assignments except for the quizzes and final course project. Too much deference to outside materials, i.e. if you really want to learn these concepts take Andrew Ng's class or read The Elements of Statistical Learning. The video lectures just breeze over the concepts and leave too much for the learner to just go and figure out. The quizzes, instead of testing your knowledge are literal the only practical learning in whole class. The course project is what you make of it.

A fast-paced course that got me going in building models and understanding the pitfalls. I felt the directions for the final project were somewhat poorly worded and vague (and calling one of the files test when it was not to be used for testing the model was initially confusing), but overall it was good. I would have liked to have seen the final project uploaded as a secure file as has been done in other courses, and Github was a poor platform for viewing html files. Additionally, the question about out of sample error caused many people problems in the projects as they confused it with with Accuracy, yet it was weighted heavily in the rubric: I'd like the instructors to review the materials how that material is presented in terms of models. I got 100%, but as always you have to pay very close attention to the rubric.

As always with this specialization, you are really just given a taste and there is no way you can fully explore all the material and references presented., but it is enough to get you going and wanting to come back and explore the material more.

Really enlightening! It has short videos that fully explicate and explain the concepts and also give you code you can use! The code actually works! I would prefer that the instructor get a proper microphone so his voice sounds not so harsh, but this can be forgiven, because the content really is awesome!

You can also download all the lectures (and there are transcripts, too), which is great for later reference.

Try it - you will like it!

Useful course that is very practical in teaching tools in R that enable Machine learning. This course is, however, not suitable for people who want to learn theoretical machine learning. For that, learners will find Machine Learning by Andrew Ng a better alternative. However, if you're interested in machine learning packages in R and how to implement them, this course achieves that purpose for you.

I give 5 stars because at the end I was able to understand the different topics and obtained a deeper sense of what can be done with the tools provided, however I believe there should be one section to have a complete exercise in order to better understand how to organize a project and put together all the tools provided. Perhaps a general diagram for analysis may be good to have as well.

I like the organisation of the course. The first video is so informative yet so simple. Great resources have been listed in it and so subtly. Also I saw the organization of folders and lecture notes and everything in Github repo for this course. It s awesome. I keep stuff like that.. well numbered and everything. I really appreciate it as it makes life of a student lot easier. Thanks.

First I want to thank very much the instructor in the online forum. He helped me a lot at the end of the course and his tutorials for gh-pages are excellent. He was also very fast in responding. Thank you.

The course did ultimately not really gave me what I was looking for. Maybe too may different facts and not enough depth. I am not sure that I can confidently say that I can build a ML model now. Technically I can, but the deeper understanding is missing. For example: When would I use which method (for example rf versus naive base), the last exercise about cross validation was not fully clear. Using the caret package is too high level for a learner. It would be better to see some more step by step examples. It was not clear to me what the expected error calculation in the last exercise was really looking at. Maybe what is missing a swirl exercises, not using caret. and then explaining how caret can simplify it. We also learned how to create a predictive model, but did not go into how the model gets updated and gets retrained, an important aspect of ML. i also do not see unsupervised learning to be covered.

lecture material could be cleaner with fewer errors

I was rather disappointed with this course. I guess it fills the objective of getting you using the caret package and getting you started with some examples. However to understand what you are doing you should defintively go somewhere else. I definitively missed some swirl exercises and more flow diagrams in the slides. It felt for me as I was just copypasting some code from the slides. The course does clearly give some good literature and places to go for details.

Project description versus requirements were terrible, not sure if the new Coursera format played a role in the issues or not. Quite a few of the homework items require guessing as the answers don't align to the results of the latest tools they have you use. If the first class or three in the series was like this I wouldn't have taken the courses.

This course was a big let-down compared to other courses in the specialization. It doesn't seem like a lot of effort went into course planning and creation. Much of the content is unclear and there is little depth. course textbook, and some swirl exercises would have helped.

I have taken nearly all the courses in the Data Science Certificate and this course was awful. The quizzes didn't work. The examples required datasets that are inaccessible. The assignments expect students to be able to do things that are not covered in the rest of the class.

Extremely confusing. Should look at Prof. Andrew Ng's machine learning course for how to clearly convey an idea.

Not enough context for the price payed

Excellent and useful course.

Some of the materials covered in Week 4 should be distributed to earlier week(s). The current Week 4 video coverage, quizzes, and the course project on accelerometer data is too much for the week, esp. if the student has lookup and review some key concepts from the resource links in the video slides. Video lectures are informative and easy to follow, although somewhat rushed in Week 4.