Welcome back to my seventh class insert lecture on Microsoft Azure Machine Learning Studio. So, we have done a lot of cross layer linear regression, right? It's time to get a new model for K clustering. And let's see here let's see the gallery actually Oh town, so it looks good. So, let's see what are there are some comments to you. Okay, good.

Um, so let's see here, how we open in the studio and see in that flow can view this and this is not last. This is, after this. I have one more class on this notebook. Because Because Promise to get you this notebook right we will. Okay, so this let's try to understand this flow chart since we already got used to what is this input data about data, clean missing data. So what they have done is remove the entire row, okay, it's up to you what you want to do.

So column in the system, but I cannot realize it right now. So we have to run it. Okay. Let's let's see later, edit metadata. So there edit the main method. They use k means clustering algorithm here because I think they are using a different they're using a different I think they are using some images.

So let's see. That's the reason the K means clustering and they are training it Training this multi class law logistic regression method, the training it again that different way and assigning to the clusters and set the columns in data sets and evaluate the model. How interesting they don't have this evaluation model for this one. Let's run this and try to see what's what's happening. Because we cannot visualize it until we learn it. So this is very good.

This has good likes and a lot of comments. People liked it. So probably it's a good one. So, so you can see actually which one is running this is completed. This is completed after completed this month, Take some time, because this is completely different method says k means clustering. So far we were only doing with linear regression and B, you know, being in regression method, and poison, no regression method, but this is a K means clustering method algorithm.

So this is a completely group ID status like, I think it has to do with images or something, probably. I can be wrong, but let's, let's see here. I've never done that should complete in a moment. Yes, good. Okay, great. So let's visualize it.

So we know what is a we don't know what is the data, which data they're using, we have the data. So they have this size, which is a string feature. Okay. column one, column two, column three, column four, this name is put some really good name, this column columns are really not so good, okay. And then they remove the missing data, there was some missing data. Okay, it's fine.

I'm not seeing any missing data, but it's good to remove anything if you even if you won't find it's good to do this because most of them you have to clean the data. Because the data may not be always correct. So now here that change this data, we're editing metadata, so that they change the names of the columns of input data, let's visualize and see what they have changed it. So that changed a label Sr f1 f2 f3 f4 and this We can limit our interesting, okay, now what they're doing is splitting the data, we know what they're doing. They're using point six, we use point 70. Now, one of the model, remember, they're using bind six, and they're training this clustering model.

Using k means clustering. And they use this data and let's visualize this. So this is what they're getting. But I also wanted to evaluate this model. And guess I can evaluate this right? Why don't have this monitor, but we'll add it.

Nevertheless. The question is, in what column you will evaluate, so the only Okay, let's run This I guess it will evaluate based on the label and let's find out and see this how this thing works out so probably I can also invalid join this and join to this but probably might not work too but let's find out for now this one so whenever you're running as to run this whole thing from scratch so that's the reason it takes time you see here this is the time it's taking to learn so we wait to this is completed in our basement are that's what I thought was the air Okay there is no school second by Miss communistic. So we have the US core model see the view the school model and then evaluate the model. Now we use this you cannot use this because it's a train model.

Okay? So cannot connect toy cluster to AI comm dotnet See this model that's the reason you won't have this. So this is still have an input for train model is uncontrolled. So you have to do this thing again. Yeah, Dream models He already have this string clustering monitor but you cannot do that it's interesting that you cannot do this this thing doesn't doesn't work for this simple reason because I'm this is clustering Mar, you need crane clustering model. So we use trained clustering model.

Okay. Oh yeah, we work on labor and I hope scorecard score model input is input right so it has to be this. So are these cannot get the school model should be. So this turn we're gonna get any graph on this one right. So you can do it any bit up. So this only graph is that this qubit here.

So we run this and this tree interesting martyring cannot have this graph Okay, so we need that one alone. And we'll see here. But Katherine murder, it's completely different right? And clustering model is entirely different from this, you cannot use a scoring model for clustering or you have to have this train model. And if you're using train model probably can still use a strain model using the split data. Right?

I can use a split data and use this, but then there is no point of using that. So that will be the cause the different data but let's try that. Okay, so here, let's visualize it. This is good. So oral accuracy, average accuracy, and you got some metrics here. So this is how you predicted this, this is good.

I mean, use this train model. Actually, this is the same thing. So there we use this one, so there's nothing the output to see this one, this is the only thing that's pretty much, or if I'm using this to be the same thing, this train model, it's called model and valid model. So there's no point, right? But they're using multi class logistic regression. That's very interesting.

And they're here they're using k clustering, but I'm not terrible to evaluate that and to get the graph. Unfortunately, I have to look around for this to see which graph can I get for this one? Okay, but for this one This is good. You get this graph and we will, you know run as predictive work. Okay, in creating predictive experiment, please get something. Okay.

Let's run it once again and see this. So you have to actually ensure stop in between when and I have to go on once you ran it, you have to set up a web service immediately. So inputting the data and then cleaning this thing actually takes some time. But this is good. I mean, this is good to know about this one. And this has nothing to add actually technically cannot add anything here.

The only thing they added here is streaming this is good group is data. So this is k means clustering and getting the data set from this and comparing that label with that and evaluated the matter which they use big data here. You can use either split data, or you can use previously we used a different thing right we can use this for some reason. Okay. So it's automatically creating new web services for you. Right?

Good. So we deployed. Okay, so let's call it this, they have created a predictive experiment. And we are brand new. We have to learn this predict to experiment. And then we will deploy this as a web service so that you can use that new API get generated and you can get the Excel graph and you can also use all right So this after this, now we'll go to one more class, and that will be the pretty much.

Okay, so let's deploy this. So that will be pretty much our last class. And you know, this is good to know, we have this thing, so Okay, good. So let's test it. So we have this one. And you can download this as xo and this will the API key generated for you.

So that's the experiments. So that's pretty much for clustering group is data, this was about clustering group is this is completely different. From the data what we have done so far, that was classified we have done linear regression, we have been Poisson regression we have been this is k means clustering. So good amount of Got some skills to bring the data. Okay, so that's pretty much for this class. We I will be back for the next class probably that will be short video for logbooks and see that and with that we'll end this you know session for Microsoft Azure Machine Learning Studio.

And I hope you like this video and thanks for watching.