You will be learning both classification and regression algorithms in detail in in the machine learning course that you might be wanting to apply for okay, so I am running several models right I have actually run more than four or five models of classification there are various different classification techniques, so I have run four or five different classification techniques I've come out and and I've measured each of them you know you evaluate the model just like how in an SDLC life cycle you test your core model evaluation is the way to actually test the test the performance of your model right for each of the techniques what is the performance right and once you once you really measure the model once you rush measure the model you start you know you get model output the output of the model is measured in terms of model accuracy it is also measured in terms of you know various other terms like AUC area under the curve we call it okay so typically in this case you know my model is 97% accurate.

This means every time I pass in a data I can claim that my model is my model can predict to 97 percent accuracy what the customer segments are okay so this is not how in real life this will be once a model is accurate 97 percent it doesn't mean the model will always be accurate 97 percent this accuracy will keep changing as and when the data changes the business realm changes the business model it is changing business rules changes but for this you know webinar you could assume that this model is 97 percent which is a very highly sought after accuracy okay in real life this is not always possible okay.

So once I run all these models you know I can also look at comparing which of my models are better is support vector model better than random forests are random for us better than decision tree, so I I create you know machine learning your Python and machine learning algorithms allows you to create this kind of you know this kind of table where you can support where you can compare different models right and say okay in all this case as a random forest is giving me the highest accuracy right and finally, whichever model that you are choosing typically in production you don't run all these models to run your production you take a sample of your production data run your models and then come out with which is the best performing model and then once you find the best performing model you run your entire production data with that model right so certain times you might want to see that your model is not working well.

You will also have an opportunity to boost the performance of your models so model boosting techniques are something that you will learn as well okay and with all of that you finally once you run your model you end up creating visualizations the visualizations would be what is your actual versus predicted visualization right you can create visualizations either in Python or in tableau typically for the management or an organization you know visualizations are done in tools like tableau or power Ba okay so this is a quickie about how a machine learning algorithm would work, okay I know I've covered a lot of topics in just probably fifteen minutes but this is what your course would probably cover in various means various ways across various months okay so um that that would be from my side I'll put back to you sure thanks Allen, so I think I guess working guys absorb about the way on in describing you world process right in the first frame yeah.

One team is obviously some of it may not have made sense to you because you know it requires some looking at the videos understanding of the topic is and then when the membership sessions happen right some of these things then start to make sense because then that's where mentors Vietnam will work with you and then take you through in this kind of detail all the topics that you have covered right, so I mean and these are the kind of things that you have to learn for to learn machine rolling in here hey so just imagine let's say so now Harlan was there taking through some of these complex topics, and I'm sure some of you might have not understood it but then imagine that if you were doing it on your own right then and going through the same topic you would definitely need a support system and that's where the mentorship comes in that's the whole idea of mentorship that you consume some content and this there is a support system from our end from the mentors that you can speak to the mentors ask questions mentors runs through in this kind of detail of course topics covered.

So that kind of then starts becoming like more guided everyweek whene some pre-work is done from your end, and than you met those folks with you two, and so they're the topics are you know you understand what are what was covered in detail like oh by the way what Anand was talking about just give you a simple analogy right he was talking about the segmentation is becoming more and more you know only used to low segmentation once in here once in one month now it is becoming more and more instrument guessing you respect it right if you go to let's say Facebook or Amazon right now you would see that the ads displayed to you are very different from let's say the asks paid to somebody else and based on your online behavior that kind of acts also changed very rapidly okay.

That's the kind of marketing that's the kind of segmentation which is happening almost on a real-time basis right and then specific messages for specific kind of segments are being shown so that's the power of machine learning which is normally harnessed to drive better business outcomes to deliver the right message to the right audience right so that's what it is that's the plaintiff news case which one was talking about hey so um one thing that happens with this I mean if you look at the pure-play online model where you have videos to consume first you don't have a support system the topic itself is a little involved as I said, and you also observe a what's typically the completion rates for online programs are only 20% and which means like four in five people attending they do not finish the program and even if they finish the program we don't know what's kind of cutting outcome have people these guys caught right and this is even lower for technical subjects.

So at great learning, we take a lot of pride in saying that our programs have more than 90% completion rates and that's one of the targets that we have for ourselves right and that's one thing that we really monitor and a bunch of team works with each participant to make this happen right so there is support from the mentors as we discussed there is a program manager which is attached to each batch and works with each student it helped in the learning journey then there's a technical resource who works again who helps with resolving technique where ease of the ratios or specific queries technically members can take it and also the technical fix it so the whole team works with each participant to provide that level of support.

That the learning outcome for each participant is met in our case it is met for more than 90% of the times and this human the 10% that you don't see here is because people have to travel on-site or sometimes that becomes a reason but most of the participants almost all the participants who both were programs actually end up finishing it with the right learning outcome so this slide summarizes some I mean we have so some of these commands from our alumni are a testament to the pool model the online best mentoring model wherein it shows that if it was really working right.

Post a Comment

Previous Post Next Post