Practical Machine Learning

Provided by:
5/10 stars
based on  14 reviews
Provided by:
Cost FREE
Start Date TBA

Course Details

Cost

FREE

Upcoming Schedule

  • TBA

Course Provider

Coursera online courses
Coursera's online classes are designed to help students achieve mastery over course material. Some of the best professors in the world - like neurobiology professor and author Peggy Mason from the University of Chicago, and computer science professor and Folding@Home director Vijay Pande - will supplement your knowledge through video lectures. They will also provide challenging assessments, interactive exercises during each lesson, and the opportunity to use a mobile app to keep up with yo...
Coursera's online classes are designed to help students achieve mastery over course material. Some of the best professors in the world - like neurobiology professor and author Peggy Mason from the University of Chicago, and computer science professor and Folding@Home director Vijay Pande - will supplement your knowledge through video lectures. They will also provide challenging assessments, interactive exercises during each lesson, and the opportunity to use a mobile app to keep up with your coursework. Coursera also partners with the US State Department to create “learning hubs” around the world. Students can get internet access, take courses, and participate in weekly in-person study groups to make learning even more collaborative. Begin your journey into the mysteries of the human brain by taking courses in neuroscience. Learn how to navigate the data infrastructures that multinational corporations use when you discover the world of data analysis. Follow one of Coursera’s “Skill Tracks”. Or try any one of its more than 560 available courses to help you achieve your academic and professional goals.

Provider Subject Specialization
Humanities
Sciences & Technology
4733 reviews

Course Description

Learn the basic components of building and applying prediction functions with an emphasis on practical applications. This is the eighth course in the Johns Hopkins Data Science Specialization.
Reviews 5/10 stars
14 Reviews for Practical Machine Learning

Ratings details

  • 5 stars
  • 4 stars
  • 3 stars
  • 2 stars
  • 1 stars
  • 5 stars
  • 4 stars
  • 3 stars
  • 2 stars
  • 1 stars
  • 5 stars
  • 4 stars
  • 3 stars
  • 2 stars
  • 1 stars

Rankings are based on a provider's overall CourseTalk score, which takes into account both average rating and number of ratings. Stars round to the nearest half.

Sort By
Student profile image
Student profile image

Student

4/10 starsCompleted
3 years, 1 month ago
I know the go-to course for this content on Coursera is by Andrew Ng from Stanford, but since that one focuses on Matlab and not R, I opted for this course instead. I should've Googled the reviews outside Coursera before I paid $49 for this (Coursera probably manipulates their ratings somehow because every course has high ratings on their panel). There is really nothing to recommend about this course. The materials are half-assed and rigid, mostly just using caret and telling you to do this and that without really helping you to understand the concepts. The practices and quizzes are too simplistic, and in fact in many situations you can't even get your answers to match theirs because of software or version differences. I actually work as a data analyst and I can vouchsafe that you wouldn't begin to be able to apply this to complex messy real-life data. I would say if you want classes on Coursera that go much deeper into the subject m... I know the go-to course for this content on Coursera is by Andrew Ng from Stanford, but since that one focuses on Matlab and not R, I opted for this course instead. I should've Googled the reviews outside Coursera before I paid $49 for this (Coursera probably manipulates their ratings somehow because every course has high ratings on their panel). There is really nothing to recommend about this course. The materials are half-assed and rigid, mostly just using caret and telling you to do this and that without really helping you to understand the concepts. The practices and quizzes are too simplistic, and in fact in many situations you can't even get your answers to match theirs because of software or version differences. I actually work as a data analyst and I can vouchsafe that you wouldn't begin to be able to apply this to complex messy real-life data. I would say if you want classes on Coursera that go much deeper into the subject matters and to learn things you can apply directly to your work, stick to Duke and Stanford.
Was this review helpful? Yes0
 Flag
Student profile image
Student profile image

Student

2/10 starsCompleted
3 years, 1 month ago
This course should be called "Using Caret for Machine Learning". If you don't already understand the basics of machine learning, you are going to really struggle. Do not take this class for any reason other than seeing some basics of how to use the caret package. This is at the bottom of the Johns Hopkins Data Science track. I wish I could earn the specialization without wasting time and money in this class when I know I am going to have to go elsewhere to actually learn the topic.
Was this review helpful? Yes0
 Flag
Karthikeyan Sankaran profile image
Karthikeyan Sankaran profile image
8/10 starsCompleted
  • 4 reviews
  • 4 completed
3 years, 11 months ago
Content - The focus on the course was in utilizing the 'Caret' package in R for Machine Learning algorithms. I liked the fact that it helped me learn that package well. Having said that, it also limited the learning to that particular package and its usage. Instructor - Good. Techniques were clearly articulated Provider - Coursera is an excellent platform for learning. Really enjoyed the experience.
Was this review helpful? Yes0
 Flag
Abhinav Sunderrajan profile image
Abhinav Sunderrajan profile image
6/10 starsCompleted
  • 1 review
  • 1 completed
4 years ago
The course coverage is very superficial. The instructor hurries through skims through several techniques without spending enough time to motivate any algorithm used. The only reason I took this course is to complete the specialization. A far better course is taught by Professors Trevor Hastie and Rob Tibshirani (in R), the accompanying lectures for Introduction to Statistical Learning.
Was this review helpful? Yes0
 Flag
Student profile image
Student profile image

Student

4/10 starsCompleted
4 years ago
The video lectures are terrible - they are too high level to understand the content in any satisfactory detail. I had to read the 'Elements of Statistical Learning' and 'Introduction to Statistical Learning' along with the course.
Was this review helpful? Yes0
 Flag
 profile image
 profile image

6/10 starsCompleted
  • 1 review
  • 1 completed
4 years, 6 months ago
The professor is not prepared well for the lecture, and most of the time just simply read the lecture notes and explain one or two sentences, the last two weeks lectures are awful, many typos and details of the knowledge is so unclear. The content of this lecture is good, but still not recommend this course.
Was this review helpful? Yes0
 Flag
athena w profile image
athena w profile image
1/10 starsCompleted
  • 1 review
  • 1 completed
4 years, 11 months ago
I am enrolled in data science certification. Thus it is not a free class for me. I expected to learn some basics of machine learning but class does not teach much about machine learning. It mostly addresses how to use R caret package. I don't think it is useful because what is the use of knowing some language syntax without understanding concepts. Just like statistical inference and regression analysis, teaching and slides are poor and are little helpful to gain any knowledge. I end up going to youtube and watching videos from stanford,caltech,etc to understand concepts. What is the use of paying for the class if I am learning everything from other resources? I have completed all courses except machine learning in data science certification track but I may withdraw from this course since material does not seem to be adding any value. Quizzes and project are too vague with little help from actual course material.
Was this review helpful? Yes0
 Flag
Geoffrey Anderson profile image
Geoffrey Anderson profile image
10/10 starsCompleted
  • 1 review
  • 1 completed
4 years, 12 months ago
**** **Course strong points: ** ** ** **This course is easily both the fastest and most complete in existence, in terms of overview all the way down to coding level detail -- for a mere four week course! ** The breadth is such that it hits all the most important concepts of machine learning which can possibly be taught to a ML novice in just four weeks. It’s a crash course for sure but you still can really learn a lot, and use these skills to to be able to solve an actual problem such as the course project. The lectures are information-dense, move fast, and are seriously practical. Thankfully the forest and the trees are both presented, and each is given attention with an emphasis on results and getting you working on actual studies as soon as you can. In fact your first study is due in just 3 weeks after the start of the course! I find myself rewatching the same lectures over and over to make sense of everything the ins... **** **Course strong points: ** ** ** **This course is easily both the fastest and most complete in existence, in terms of overview all the way down to coding level detail -- for a mere four week course! ** The breadth is such that it hits all the most important concepts of machine learning which can possibly be taught to a ML novice in just four weeks. It’s a crash course for sure but you still can really learn a lot, and use these skills to to be able to solve an actual problem such as the course project. The lectures are information-dense, move fast, and are seriously practical. Thankfully the forest and the trees are both presented, and each is given attention with an emphasis on results and getting you working on actual studies as soon as you can. In fact your first study is due in just 3 weeks after the start of the course! I find myself rewatching the same lectures over and over to make sense of everything the instructor is saying because there is a lot there, and it’s explained in plainest possible language given what the subject is. There is little wasted movement by this instructor, and pretty much every word means something important. The instructor obviously knows this material very well and I think he applies machine learning himself, for a living, in real life. However please notice that the instructor’s language is plain and simple, and he is definitely not trying to lose you. He is trying to keep you going and understanding by whatever it takes, like a master. He knows who his audience is -- beginners -- and you can usually follow what he says if not the first time, then the second time, or the third time. I am very glad I took this course. I feel like I can actually do machine learning studies, or rather, I know I can, because I actually did. The course project is proof. The course project was not trivial either; it is not a toy problem. It was actual data from actual field studies of wearable accelerometer data, and it was not reduced in volume or complexity in any way from what I can tell. It was pretty hard to solve at first. Eventually my solution worked and it worked very, very well. I know it worked well because I verified its accuracy with cross validation and out of sample error measurement, as well as on predictions using the 20 graded input observations. And I am ready for doing more courses in this subject, as well as doing other practical applications of machine learning and predictions. It’s great to get a feeling that you can do this, and know that you really can. I literally can’t wait to do more predictions. I give a sincere and big thank you to the instructors at Johns Hopkins for creating this course and the course sequence it is a part of, and letting me study your materials. Course weak points: In summary, quizzes in which careful assessment of student knowledge has been given a lot of thought, do certainly exist on Coursera. This course is not among the best of them, unfortunately. There are certainly some courses on Coursera in which accurate of assessment of student knowledge has been given a lot of thought on the quizzes and exams. To be clear, it is actually very good to allow many attempts at solving nearly the same quiz question by students, because this can drive the student to do additional review and study and research into the material on that particular question. Similarly, the grading rubrics for the course project sometimes give the impression of being of rushed quality, as if not enough time was put into designing the grading criteria; or there are not enough criteria on which to grade students; or the criteria seem to be overlapping when they should be distinct from one another, leading to confusion by the grader, and high variability in the same student’s scores I expect. This is seriously a lot of material to learn in a little bit of time, that is, if you actually want to understand it. There is only so much machine learning theory and practice an ML novice can possibly assimilate in their brains in four (or 3) weeks, even when instructed by the best. As a consequence it took me quite a bit of additional time to digest, think about, and research the material covered in the lectures before I felt like I decently understood what is going on, and understood what to do next, in my machine learning projects. Don’t bother attending if your programming skills are weak. You must program in this course, but without spending time on programming, if this makes sense. That’s OK for me, but this is a warning if you are not confident or ready. The course sequence includes an R programming course, given early in the sequence before you get to this ML course. Make sure you can pass the R programming course before taking the ML course and you will do fine. Unfortunately the instructors don’t participate in the forums much or at all, from what I have seen. There are forum teaching assistants and often they are very helpful, so that's good. But, the forum’s volunteer teaching assistants don’t really have much skill yet compared to the instructor -- often they are merely past successful students of this course -- so it’s sometimes a case of the (at least compared to the expert instructor) blind leading the blind despite everyone’s best efforts. There are relatively fewer students per course offering and this works against you in the forums because there are not many people who know the answers or can discuss things with you. Here’s why this is the case: This course is re- given every month. I took like the 6th and 7th offerings. Also the course is about the 8th in the sequence of Johns Hopkins Data Science Specialization. Naturally, few people have the persistence to make it all the way to course 8 in a sequence. As a result you end up having probably about 1/100th of the number of students participating in the forum for this course versus some other Coursera courses I’ve been in, and this is why I say there are fewer participants in the forum. There is no homework, so in this course, the weekly quizzes become your primary hands-on learning tool where you try out machine learning yourself for the very first time. Also notice that nobody is permitted to share the quiz solutions. What happens then, predictably, is that all too often you end up never figuring out how to solve some (many?) of the problems. It would definitely be better for learners to have some help being shown specifically how to solve all the quiz problems we got wrong. Perhaps adding some homework assignments, in which students are permitted to share solutions might be a way to help everyone learn everything. That said the course workload is already quite high especially with the challenging research paper, and the four quizzes, and all the lectures, and I don’t know how you would add homework in the mix too in four weeks unless students work 15-20+ hours a week on this course. The so-called 3-5 hours a week is definitely NOT the workload now, even without any homework, despite what the syllabus claims (unless you already knew ML, or you don’t care to understand the subject). Maybe you can pass on less than 15 hours a week but I could not! Generally, you can pretty well see what is being learned by other students in general, by looking at a smallish sample of research project papers. I just finished peer grading about 6 papers. Two or three of the papers seemed to be suspiciously near-identical copies of each other in terms of showing the same background, nearly the same data filtering steps, same data partitioning, nearly the same variety of models claimed to be evaluated, but were not actually explicitly evaluated, same omission to compute the error on in-sample (training) data, same choice of winning model of random forest, same list of non-selected models that were claimed to have been assessed but for which no code evidence was shown nor accuracy percentages discussed, same claim of use of cross validation, and the same omission of actually showing the use of such cross validation code despite the claim of its existence. They all also chose to parallelize their code instead of letting it run sequentially, and they all achieved parallelization using the same three function calls using the same nesting structure in one line of code. What’s the chance of that happening with honest original student effort, when there are so many different ways to do a machine learning solution? I think the chance is super-duper low to have so many similarities. And these students had very little discussion to add beyond just code, which is strange since you would think anyone who spent time assessing four or five different models, with training and testing data sets, would have had a lot more material to discuss, and some specific accuracy numbers to share, but no. Other papers started strongly, and originally, but just never finished, never showed the prediction output or accuracy of everything above, as if the code perhaps did not actually execute. There was no evidence that the code executed via a print statement of the confusion matrix, nor at least any discussion of the results that such code might have produced, but which the author simply forgot to put in the report. Consequently I don’t think the prediction code actually executed, or predicted anything at all, accurately or inaccurately. Maybe such people ran out of time to work on the project and just submitted whatever they had, when they ran out of time to do course work. I wouldn’t be surprised. The proposed 3-5 hours a week of course work was not enough for me, so I spent quite a bit more than that, about triple that. Other papers had authors who did not seem to understand why their prediction design worked so well on the training set, but then did not work very well on the 20 graded test question predictions that were also required to be submitted. Well, that’s ALL about the central idea of testing your predictions on training (in-sample) data as well as running predictions on testing (out-of-sample) data, and comparing the two different accuracy values that pop out. It’s an essential concept in machine learning, generalizability. They had no idea, or if they did, they completely forgot to discuss this issue in their study, and how it applies to what they just saw in their own model predictions. **** **This course is one of a kind, and overall, this course is excellent. I am super glad I took this course. ** There is nothing else that can get you up to speed on a pretty challenging subject like machine learning, in such a short time -- not any other book, not any other course. It is a perfect introduction for a person with ambition. ** ** **That said, this course is definitely not enough on its own. You really need to take some additional ML courses, and read some more books on the subject. This course gets you on a good start and ready for more. **
Was this review helpful? Yes0
 Flag
Leonardo Dos Santos Pinheiro profile image
Leonardo Dos Santos Pinheiro profile image
6/10 starsCompleted
  • 2 reviews
  • 2 completed
5 years, 3 months ago
The topic is interesting and truly best covered in other courses. Intro to Statistical Learning, The Analytics Edge, Andrew Ng's Machine Learning, Learning from Data and others provide a deeper insight into Machine Learning. If you have taken one or more of these courses you could still benefit from learning a little more, specially about the caret package, but if you're new to the topic you probably won't benefit much since the lectures are brief and too fast paced. In summary, if ou already have some experience with machine learning you may benefit from seeying some new topics. But if you're new, you may not really get enough knowledge about machine learning.
Was this review helpful? Yes0
 Flag
Jeff Winchell profile image
Jeff Winchell profile image
1/10 starsCompleted
  • 91 reviews
  • 66 completed
4 years, 10 months ago
The Johns Hopkins Data Science series is a mish-mash of often poorly designed assignments and typical last-century style lectures. This class is no different. If you want to get the answers right, you will need to spend a lot more time than the advertised 4 hours/week unless you are cheating/gaming the system or already know this topic. The quiz questions are often poorly explained, so you'll need to look at the forums to figure out what is going on. The professor is 100% vacant from them, and there was no TA present at all either. Given the problems with the assignments they created, this is inexcusable! They are trying to run all 9 Data Science MOOCs this month, so the staff is spread too thin. The bright side is that by struggling so much trying to figure out how to answer the quizzes and project, you will very likely retain that hard-won knowledge longer, assuming you don't drop the course out of frustration with its poor impleme... The Johns Hopkins Data Science series is a mish-mash of often poorly designed assignments and typical last-century style lectures. This class is no different. If you want to get the answers right, you will need to spend a lot more time than the advertised 4 hours/week unless you are cheating/gaming the system or already know this topic. The quiz questions are often poorly explained, so you'll need to look at the forums to figure out what is going on. The professor is 100% vacant from them, and there was no TA present at all either. Given the problems with the assignments they created, this is inexcusable! They are trying to run all 9 Data Science MOOCs this month, so the staff is spread too thin. The bright side is that by struggling so much trying to figure out how to answer the quizzes and project, you will very likely retain that hard-won knowledge longer, assuming you don't drop the course out of frustration with its poor implementation. In medicine, there are only a few institutions in the world at the level of Johns Hopkins. Unfortunately that is not so with their computer science department and I can imagine their actual courses even in that department are far better than the MOOCs they put on at Coursera. I would think their administration (and marketing people) would not be happy if they saw the Johns Hopkins brand being sullied with such poor courses that are widely available because of the Internet. I only give this class a 2 because at least the professor isn't as poorly organized as the one teaching the Regression and Statistical Inference classes. I only continue in these courses, because it is a way to get me to learn this material better (and I'm almost done with all of them). If you have the option to take better courses in Machine Learning/AI/Data Mining (i.e. from Stanford, CalTech, Berkeley) take it and skip this.
Was this review helpful? Yes0
 Flag
Richard Taylor profile image
Richard Taylor profile image
1/10 starsCompleted
  • 29 reviews
  • 28 completed
5 years, 4 months ago
Machine learning is a very hot topic today. This course aims to teach how to use ML algorithms using the caret package in R. Caret is a wonderful package that is only scratched in the lessons, there are several fantastic tools in caret that are not even mentioned. There's also very little about how to apply machine learning algorithms in real problems or advice about how to use caret for real world projects. The course has 4 quizzes that are easy to solve and then a final project where you are given a dataset and you have to apply ML to predict values of a test set. Unfortunately the logistics for the project are horrible and the way it is graded is really bad. This is as bad as a course can be because it covers a very interesting topic a very interesting tool and yet it fails to be entertaining or even educative. A complete shame.
Was this review helpful? Yes0
 Flag
Hamideh Iraj profile image
Hamideh Iraj profile image
6/10 starsCompleted
  • 70 reviews
  • 60 completed
5 years, 4 months ago
Up to now, It is the second best course of data specialization series (after developing data products). I am somehow accustomed to the ways these 3 professors teach and found a way to learn. In comparison to data analysis course, it is completely redesigned, very much improved and many tips are added to the course and the use of only one package (caret) brings the tool usage into order. Remember that it is a 4 week course and you cannot expect to learn the wide variety of concepts of Machine Learning and it cannot replace machine Learning by Andrew Ng which is far better in concepts. My recommendation is taking (or auditing) the Andrew Ng's course (you have to work with MATLAB or Octave which I did not like) and this course as a complementary to learn how to work with some R packages and how to map ML concept to R programming language. Probably a good idea is to extend this course to two or three four-week courses on the series. st... Up to now, It is the second best course of data specialization series (after developing data products). I am somehow accustomed to the ways these 3 professors teach and found a way to learn. In comparison to data analysis course, it is completely redesigned, very much improved and many tips are added to the course and the use of only one package (caret) brings the tool usage into order. Remember that it is a 4 week course and you cannot expect to learn the wide variety of concepts of Machine Learning and it cannot replace machine Learning by Andrew Ng which is far better in concepts. My recommendation is taking (or auditing) the Andrew Ng's course (you have to work with MATLAB or Octave which I did not like) and this course as a complementary to learn how to work with some R packages and how to map ML concept to R programming language. Probably a good idea is to extend this course to two or three four-week courses on the series. students expect to learn more about the core topic of the series.
Was this review helpful? Yes0
 Flag
Student profile image
Student profile image

Student

5/10 starsTaking Now
5 years, 4 months ago
Expected to learn R and machine learning. But now easy to follow the speaker. They do not provide the time for the students to think.
Was this review helpful? Yes0
 Flag
Greg Hamel profile image
Greg Hamel profile image
6/10 starsCompleted
  • 116 reviews
  • 107 completed
5 years, 5 months ago
Practical machine learning is the 8th course in the 9-part data science specialization offered by John Hopkins on Coursera. This course introduces machine learning in R, including the basics of prediction, splitting data into training and testing sets, regression, trees, random forests and boosting all in the span of 4 weeks. The course focuses on using the Caret package in R to apply machine learning algorithms. Similar to other courses in the data science specialization, the course content is mainly static slides with voice- overs, but thankfully the slides are generally not overly cluttered and the voice-overs are of decent quality. The course has a lot of good information on how to use R to apply common machine learning techniques to data, but you aren't going to gain a deep understanding of how the machine learning methods work. "Practical" in this case means "learn how to use the tool, not how it works." I suspect students comi... Practical machine learning is the 8th course in the 9-part data science specialization offered by John Hopkins on Coursera. This course introduces machine learning in R, including the basics of prediction, splitting data into training and testing sets, regression, trees, random forests and boosting all in the span of 4 weeks. The course focuses on using the Caret package in R to apply machine learning algorithms. Similar to other courses in the data science specialization, the course content is mainly static slides with voice- overs, but thankfully the slides are generally not overly cluttered and the voice-overs are of decent quality. The course has a lot of good information on how to use R to apply common machine learning techniques to data, but you aren't going to gain a deep understanding of how the machine learning methods work. "Practical" in this case means "learn how to use the tool, not how it works." I suspect students coming into this course with no prior knowledge of machine learning will find that the lectures jump from one topic to another too quickly as the course goes on. Taking a course that covers machine learning theory, like the 3 part machine learning series from Udacity, will give you a deeper understanding of the methods introduced in this course. Practical machine learning does pretty good job introducing a machine learning topics in a limited amount of time, but the coverage is too brief to gain a solid understanding of many of the methods presented. This course would have been much better if it was 8 weeks and had at least 1 hour of solid lecture content per week with interactive exercises or homework. If you’re looking for an excellent practical machine learning course that spends enough time on each topic and has enough homework to really help students learn, check out MIT's Analytics Edge on edX.
Was this review helpful? Yes1
 Flag

Rating Details


  • 5 stars
  • 4 stars
  • 3 stars
  • 2 stars
  • 1 stars
  • 5 stars
  • 4 stars
  • 3 stars
  • 2 stars
  • 1 stars
  • 5 stars
  • 4 stars
  • 3 stars
  • 2 stars
  • 1 stars

Rankings are based on a provider's overall CourseTalk score, which takes into account both average rating and number of ratings. Stars round to the nearest half.