Welcome to my fourth class for Microsoft Azure Machine Learning Studio. So here in this class, the class idea is good Tom. So we have completed prediction of income. So that was kind of automated tutorial till it was automatic. Now, what we'll do here is we will create our own and guided this we created an algorithm right? I mean, of course, it was tutorial kind of thing and we created Web Services now.

They can delete this, but but we cannot delete directly you have to go to the Delete first web services, then you have to delete this one. So let's let's let me show you something actually, right now you can add this I think you can add this to the project. Let's go to this Udemy and let's see Add that one in the project. So that goes to the ad set. See this experiment to see this you see this is there are two expenses. One is the income prediction.

And the other is. The other is some predict Do you know is Web Services actually go ahead and click on that. So, three things got interesting. Yeah, so that data set and click on it. Now one good thing about this here is still have the people who have access will view this okay. So that is only the use of this project, actually.

So, here you can see the two experiments. Actually one experiment, one data set and one prediction. prediction. Two experiments of course, that's two experiments because one we use for web services because Web Services manipulated limited Got that, um, but I guess on this one for this. Okay, so here, let's use an another experiment, we'll go through this two more experiment because I'm sure this will take some time to understand it. This is not easy.

I mean, the whole concept of machine learning is to machine learning. I mean to, to learn a machine right? To predict our machine, and futures. So kind of prediction are not for us for our own purpose. But for example, let's take an example for drove. You want to clean a drone.

Let's say you you're renting a drone, right? And you are, let's say, let's take an example. You're learning a drone without a map without manual interference. That means no one is interested or this drone is automatically moving. So you want to make sure so There with machine learning is coming to the picture, you develop an algorithm for that machine, that if, let's say the guy who wants to crash, but it doesn't crash, because there is an algorithm that if there is anything in between, it moves away from it. So that's the inner.

So that's, that's the machine learning. Okay? I mean, I'm giving an example. That's just one example where you did just the machine learning application some of why you wear it, you know, you have you can use it you use in healthcare projects, commercial projects, or industrial private military application, oh my god, this is larger than the applications. But here we are using for very, very lenient portraits, like for potential income for production price. Let's, let's do one more experiment.

Actually. Let me actually, I'll use a bit of document here. So I mean, The documents are there for you. Okay. It's called shaming. Um I have included this in this 10 MB of documents so you can go ahead and view that one let's do this.

So I am actually you know, so this lot of things you can view this just use a lot of information which I actually have not even discouraged about it. Right So here you have the six you get confused of my video tutorial uses tutorial and Metro to understand this. You know, we go through this I wanted to show you another one but unfortunately, it's like 350 pages, use it. I hope it will be useful with this video, use it with first you know, understand this video why we are using this machine learning Azure Machine Learning Studio and then use it to Common documentary that will make more sense, or vice versa, it's up to you. So here to go for issue machine learning so good. Okay.

So, now what I do here is so good, okay, no, we see here, I will go back again in the experiment section, they have this lot of experiments, we are creating experiments so far we created all the experiments. So, let's go back to experiment section. Go ahead and click on new sample, we'll go through the sample if we have time, probably for me for anything tomorrow or something was somebody three, but two examples we'll try to see and we'll also we have our to do at least five hours, five or six hours of training at least you have some idea of this. We'll go with this notebook and see if it works. Not try that but let's see what happened. Okay, so go ahead and click on Blank experiment, okay good.

Now when I see this blank experiment, right, you this thing automatically shows up none of the save data states here is to solve the things showed up. So I give it a name here as automobile, automobile price prediction. All right. All we are using is flowchart everything is into chart. Okay. So what we need is data, data, data data always it's a data, okay, so we're not going to use our combined price data.

Did that guy here and see here. Okay, grab that guy. Put it here. That's it. Simple as that. Put auto by there's too many But you can sample so you want to get data.

This is data data set right now. are now we tried to visualize it, visualize it, click on number one, this global the number one in con, visualize it here, if you see this on this. There's a lot of things here. Okay, the two, not five rows and six columns. So this is the data actually supposed to be in database. This data is supposed to be in database, but here you're showing up in your that's a good thing.

Okay. So there is a lot of data here, right, not of data. And now, I'm looking into price. Let's say I'm predicting price here. In for the ease of convenience, you can do anything I mean, you don't have to predict only price. You can predict horsepower, you can predict compression ratio can put a stroke.

I don't know you can put it anything, whatever you want. There's a lot of columns, 26 columns, you can predict all of them. I don't care. Right? It's up to you. We are doing machine learning.

So you want to predict what what may they be? What's the prediction ratio, I'm sorry, words, you want to split it by train the model. First thing is you have to train the model with a 75 or 70% of 7% of data. And then you have to visualize the data. See that how how it looks like? I mean, it's nothing to do with what to do with it.

It's for machine to local users, but you can view this graph. It's very important actually, to be honest. Probably, I would write a paper on this papers, so many people, probably I will write a paper on this machine learning. Right? It's that complex. Okay, so you can predict fuel types.

I really want you to roll this Gas diesel. This is pretty. I'm not aware it does not dispatch during the meeting within 30 hafting splatter gas types, I don't know he's only buy gas and diesel is meeting. octane is a lot of things. octane on 10 I think have been just as many things, but they don't have it here. So insufficient data.

Okay. Now my Oh, I predict price in this case. You can predict anything. But in this scenario in this tutorial, in this class, we predict price. Data is homework. If you aren't predicting anything else, it's up to you.

Whatever you can predict. So, here I see this price. But the problem is there is a problem here. There is a problem here. There's a missing data. You see there is no data here.

This for this one. There is no price. Right. is already here. So we don't want that. Okay?

No, oh yeah, you can see the statistics. For this there is a missing data, if you see here is two data God. Only two data is from whatever is no data for that. Now, I have to be careful in that. So I want to protect data, but I have to clean the data first. Right?

So what we'll do here is so we have seen this right? So I want to select columns in the database, see, select columns. But we are doing data manipulation here. Okay. Let's click it here. And what we do here is, yeah, we want to coordinate Want to do here is I want to clean the data, right?

So what we do here is launch column selector and go the rules and exclude normalized losses, right? Okay. Just wanted to make sure that if you make some mistakes, it works, but it doesn't work as per your requirement. Okay? Now, what I'm doing is I'm excluding the columns, you know, we normalize the losses. And what I do is run this.

It's better to run always each flow chart if you run, so that shows you if there is an error in this or not, so there is no error. So we are good. Okay, that's good. Now what we'll do here is no Wi Fi The missing data, the point here, the missing data. So, click on clean the machine data and join this here. Alright, now I have this option.

Alright. So what I do here is I have to clean the missing data, right? So, there's an option How do you want to clean it? There's a lot of option replace using mice free customer substitution value, does it mean there is no value? So what I'll do is remove the entire row and last time I did some mistake I remove entire column which was bad. So remove my Tarot.

So I removed the entire log but Let's see. Okay. I did some mistake here. So, I selected our columns. Okay, looks good. So we are good at it.

We don't need to select all the columns. I want to remove the entire row, whichever doesn't have any. So we are cleaning up data here. Okay. We move this once. Now we try to run it again, right click, go ahead and right click and click on Run select.

So it's always better to run one by one so that it gives an ease of convenience of you know, for the next flowchart to prepare. Okay, so now we'll clean the data. How do we know let's find out go ahead and click on that. The two in their two bags here right cleaned dataset and cleaning transformation. We go Clean data set and visualize. Now if you see that I wanted to see the price, right, if I see the price, the price there is no longer here anymore.

Let's find out. Let's see the price you see here. Now this data is cleaned. Because I see here's the thing. You can see here's the thing, it's not necessary you have to put it here, you can remove entire row or put the missing value with zero or whatever when you want you can do that. Okay?

And replace it with you know probably stick our you know who want to put in some 00 Okay. But since there is the null I want to remove that in this tutorial in this tutorial, but you don't you don't have to do That you can append some values or based on the not upon experience built on your knowledge, you can do that. But here I'm just removing it for duties of convenience. All right, so Okay, good. Now what we'll do here, we ran that one successfully, you run it, and let's do this. Now.

Here, I'll set it again column data set. All right. Now the clean data set, this is the clean data set and I will append to this one. Now what are you go here and launch column selector? Okay is good with rules, I columns. And it's no columns and include columns.

Okay. So this so this is the important thing. So this is Work you want to show up in the as the columns to be honest, I know I'm putting all of them to be honest and put everything in symbolizing me. I don't care. But you know for tutorial point of view we will just use some of them as make body style. This is the body style of the car.

Okay. We we base to, I mean it can be anything. I mean I'd be using everything actually to be honest. engine size or the size horsepower. To be honest this is this makes our for me a little. I mean no sense at all.

Because this is what we'll be showing up but to be honest, this this is a nice talk. My wife actually But price prices impact this is what this is important if I don't care anything else, but price should be included because that's what I am predict predicting about. So click on checkmark and run it again. Let's see what happens. So this string refloat on accident each and everything and you've successfully ran really good. Now what happens here is not what you want to do is now you want to select a lot of them, but you're actually going to split the data right before selecting an algorithm to split the data.

Why you want to split the data because you want to train your model and you have to train your model based on you know, you know based on the percentage here 75 or something, right? Oops 75 right point 75 that means my percent of this data will go to this right. Now what we do here is, here are what we use last term we use to, to base by linear right. Here we use linear regression. Okay, so this is the era. This is the algorithm of linear regression.

You can use base, Bayesian linear regression to this is a radical I mean based on this, it's beta 0.001. And you can online gradient descent lottery I leave this thing alone for tutorial sake. But as a homework, you can try it yourself or what he does, right. So now we are splitting the data Now printing the model. Now here we are doing some work. All right.

So here you have two options right. Now I didn't ran this one but let's put it here. Let's put this here. Okay, so what do you want here is you want price. So delete this one. Delete this one.

Let's Let's, I know run this you have to run this because it's better to run each and every flow before you're doing this. He don't have to apart, but thing is, it's good good to know because so here I'm using a PWM algorithm called linear regression. And I'm using a train model to train it with 70% of data. So I've used this this is an untrained model and is putting into linear regression and the split model did this. Okay, that means 75% of data is going to train this. Now, I will do this launch column selector.

And here what I want is price you can sell it only one not many, because you want to train only one right now, no, that's good. So, this is this is good we are. So this is what the action the actually the whole action starts from here, actually linear regression. What we did so far was cleaning of data and this is a great raw data. I'll explain this again. We'll go through this later and explain what is it all about.

Now with the train model, run it again, run it to school run, because it will run that it takes that linear regression that 0.001 and trace the beta of the price, which we splitted in 75.75. And it shrinks the model. And we integrate some data, again, generous data build on that data. So we finish Okay, good. So once you've trained your scorecard, right score model site, so that you can see what happening, the score model actually shows you to see what's happening. That 30% should be here, I guess.

Right now We'll run this data I'm sorry, run this flow. So we are not done anything actually here we just created a flowchart. And now once you have run it, try to visualize it. We'll see what's happening here. What happened to my income website which is price. Now, the price you see here, see this price.

Now, it has labeled better the straining on your new readers using linear regression algorithm, we have predicted the scored labels 10285546 8.84. We have predicted this right? See, as you can see, we have limited columns only nine candidates may participate because we selected this one right. So technically For me this I don't care. But I will. I don't care what column Sure.

This is like the only thing you wanted to see. That's the only way. If you won't use it's not necessary this this flow is not necessary technically. But this was used to see what columns you want to see in your score model. That's the only point but technically doesn't make any difference. That thing doesn't make any The only thing difference makes it Subaru.

Okay, this make score label is nonzero grades, it's fine, Mitsubishi, it's 5446 Prime's by the side. All right. So that's how the score model. Now what we'll do next here to view this graph, now we'll evaluate that model. Right. Let's see well with that monitor, and we'll go ahead and click on that.

Run the flow. Will you just run the flow and see what happens? Okay, good. Now we have evaluated that model. Let's see this visualized. Let's visualize that, right click on that small button.

As you can see, there's a graph generated for you this error histogram in the mean absolute error. Okay, let's try to see what is exactly this means. So mean you have this option right mean absolute error is 1656147. So this is the average value of absolute error. An error is fake, predicted value on the ad and the actual value, so you want to predict a future prices of auto bytes then this is working you'll be using okay? So this shows this error, and this is a frequency.

Okay? The error in the x axis and the frequency on the y axis. That's good. And there's a root mean squared error. So normal average squared error, the prediction made on test data set, or since we use 75% so it depends on that do we splitted the data into point 775? You don't have to be fine seven, five, you can use find 5.658 or find one.

It's up to you. How do you want to this is actually a fraction of the rules what you want as an output to be great. Want to get trained, right? The machine has to get trained. So I'm using like, point 5.75% I think it's a percentage point environment. 75% I think that's what I am assuming he Let's see here are related absolute error, this average absolute error related to absolute difference between actual use an average of actual.

Okay, great. And the latest Quadir use the squared error related to that's good. So, there is coefficient of determination Where are your squared early this is statistic metric indicating how well a model have been studied. So this is how you can, you know, visualize it okay to visualize it, use this small icon located here, right click and visualize it. And you see this graph and so technically this is meaningless for you. If you see this, what's up on this URL, get it right.

And we do. I'm like what, what I have to do with this data now, this is important. Once in Because since it's a machine learning, this is for machine learning not for you. That means let's take an example for as I told you, right drone, let's say you're using a drone, and you're making drone to predict some that say you're making drone to predict without interference. It has to, let's say, MSM guys, right? They are shipping.

They made some new that they will be shipping all the goods to drones with no no manual labor's right, they'll be shipping the goods. So now they have to use this machine learning so that you do so that's it, there's a window and there is there's a window in opposite direction which the drone has to go right. Um, but there's a package to be delivered to that particular home right now. Now you have to predict that in the wind is coming into that direction. The drone has to mimic a movement to the end of the direction to avoid that when that's how this machine learning is useful. Doesn't make sense.

Right? It's, it may not be useful at this point of time for you. But when you're implementing into some system, including I just give you an example of drone, it's the front have to be drone. It could be anything. But we are using algorithm for predicting anything which based on the data, so let's say let's assume that the beam radiation or the wind and that California is always in that flow. So drawn as to predict that at that time, the green will be high, so it will not move at that time and moves away from that direction or takes a different droop.

Or, you know, based on this model, right. So this model helps you to machine in order to learn that machine that when the wind is blowing that it has to move into different ash. So that's our, that's it about all about machine learning. Okay. So, but for now, you might not understand those data because the data is a little bit you know, different is nothing to do with manual interference. It is used for the machine, since it is for machine learning, right.

So, so let's try to understand this flowchart. Let's recap, let's revise once again, try to understand what we have done. So, so, you know, what we have done is we have to can learn data, that is automobile price data, and you can visualize this. We won't see this works for data. So, in this poem, I wanted I wanted Price, but I said is that the missing data is here, two to three missing data. Two are supposed to here and one year.

So what I want to do is to clean the missing data. So I selected this, use this normalized classes, that's a different column action, exclude that column and clean that missing data. And I removed the entire row, so whichever the data is null, so you can append that with zero, that's okay. I mean, you can clean the cleaning mode. You can append that with zero because since it's a null, I removed the data so remove the entire row. Okay.

And then then what I did is select number of data for me this, this makes no sense. I could select all of the information like make body straight, don't it's Not necessary, but he is important because I want to see only the make and the body style itself I want only those information not every information I selected particular columns here see here and select only particular columns that means only that price is of course, important surprises my priority. So this should be there right in you know excluding that what else information do any do any the make body style meal, it's it's based completely on you. How do you want to look at the data act? Okay, so that's what I selected the date. And then I use from here the action starts splitting the data may want to split the data.

And how do you want to split the data in point seven, five. So here we use an example of point seven, five. And, you know, we split the data and train the model using a lean regression model, okay, we create a linear regression model and put it into this and untrained model, and then put the split data that spines and fine and trained it. And with that, we want to see the prediction using score model score model helps you to predict the future data on you know how the based on that data points are in file information of fracture flows, it predicts that score model, and then evaluate model shows you the graph, you know how the, let's visualize it again. Right, let's visualize it, visualize it and see this the only histogram we have. So here only the error with frequency.

So what could be the error if we could predict in the data that's evaluation Okay, so that's pretty much on this one. In that not pretty much uh, we have, we still have to do here something. Now we have to deploy a setup as web service, we want to set up as a web service. But since it is disabled, because we have to run it, I think we have to save it to do that. So let's save it. And we saved it.

And let's run it. Now, this still runs the whole the model works, what we have done so far it's done from scratch again. It's good. There's no error, and there's no error because there's no code here. So technically, there will be no error, or Yeah, there could be if you don't specify any data, information or anything here, there could be errors. Now, I will use set up web service Good predictive web service.

Okay, so let's see. So it automatically, you know, changing the data. You could add some of this if you want, but Okay, so creating predictive experiment was cool details. Okay, let's prove it. So now we will. I think it's done right.

Okay. Now Okay, now we want to run. Let's run this web service. This is a predictive experiment. We just trained it. This is a prediction experiment, which is the web services we know which is, which was designed with web services, and it created its own.

Do you notice it changed It's all values, it moves to two columns. It changes everything. And then you know, it takes some time here. Okay? Now, once a predictive experiment is successful, we know it's been successful. And we deploy into web services.

Now this is successfully deployed in the web services. And now you can see this API for this application program interface. This like Previously, we have seen it. You see this API, and you can test it. So we don't have much of this information for some reason. And oh, is everyone is empty okay.

So descending ran successfully but because goodness this is when someone says so. So anyway so this is supposed to run it you know I am kind of different for some reason, I think, but you can add this apps you can download it and see this. Okay, how does it looks like I have never done this let's, let's try to download and see what's exactly use this. So this is actually this will be a web service actually. It will be in web services in XML format. And we will see what is data Okay, so this is how you can you know, that's not interfere it doesn't interfere.

There's so many models you can create with so many algorithms. This is just an example a tutorial indicative tutorial for a second tutorial for for machine learning how you can use this model for machine learning, okay, so there's some internal so don't worry about it. There is XML from my site not from anything else so there will be an error. So don't worry about it. This will not show up for you. This in my machine actually some Excel was not installed properly.

So So here's the URL for this and the access key and the schema, the macros and make sure you enable the macros Okay, good. Yeah, we don't we didn't get this or anything we can 00 for some ways I'm doing tonight but yeah. This predict Oh, we can we predicted only price British price should have shown up but okay. So but this is how you can this is like web services when you're using it. So, okay, so that's how, you know you can use this one Okay, so that's about big action you know? So that's pretty much about price prediction for an automobile so he didn't do right income prediction and automobile price prediction.

I wanted to look and see this two web services to and there is one more tutorial actually predict to revolution for German actually German card. This I need to do this one to show you, but you take it as homework. You have this in this one it's it's easy this big app to edit metadata a little bit and see which columns we want. So this was not running properly in my system. It was taking too much time. Probably my internet is not good are probably because not So some reason, but Okay, so but the whole concept is, are you ready to use the R code?

Use this R script and you have to run it. Okay. And then you have to. Okay, we can use that. So let's go back to our experiments and see what's happening. So the whole point is you have to crane the data.

I'm sorry, I'm not trained the system frame the monitor, okay? This whole this, this whole thing, the start thing is actually cleaning part of your data. Let's say your data is not clean or your data is not appropriate. You have to make sure how do you work data in this the whole The second thing is here where the action start from linear regression is a you know, there are a lot more than linear regression as I told you, right. Oh, this filters you can use filter and fire filter. When the filter you can use the median filters, manipulation, so if you're familiar with ETL, right?

To mean something, either Informatica or data stage, I've never worked. And then to be honest, this will be a piece of cake for you. So you will know what data to use. So to be honest, you can predict anything so and you can clean you can edit meta data, you can join. Let's say you want to join the two data into one let's say you're using two tables speed. This is Do any data base actually, the data what you see here let's let me show you again.

So whenever you're seeing this kind of database right this is the column and don't this is columns and do not file. So, these are these are the database actually. So, you can do if you click it it has more information feature numeric feature and this is numeric feature. So, if you want to join something with you can also do that too Here it makes more complex flowchart becomes more and more complex when you try to use more. So this is so we sorry, so we keep it simple as a you know this tutorial binder, but you can do lot more thing can join the data. So SQL if you know SQL you can use SQL transformation.

Okay. write SQL query that you You want some you know, some SQL you want to know some SQL right? So there is inline view sub query nested sub query. So sometimes you want data inside the data. So, what you want to do is you want to write a inline view to the particular data which you're looking for, and only that data should be used. So, you want to use SQL you can do that too.

But that will be mostly will be before this, this part becomes more and more complex, the upper part will more and more complex, but the lower part is Ruby similar not same. Because you can train one model at a time, one model at a time, probably can do multiple, but that will become more complex, right? More complex and you do but this will make it simple we did for linear regression. Let's see here. You can do that for no Bayesian linear regression. But, you know, we have we have seen this earlier, right?

Let's find out again, let's, let's review it again, we'll have some idea, vision forest regression, you know, ordinal regression. For this, we don't require all this because it's really just a simple price prediction. But if you're doing some complex thing, you want to use those things. For now, we've been doing in Congress. So that's pretty much for this. Okay.

Let's, in the next class, we will go, you know, probably for me will be tomorrow, probably. And so we go to this. We've set we use some of these samples. And see try to see this. You know, and and view those Those guys flowchart they'll be more complex for flowchart. This is where we created it simple ones very simple.

It's just a joke. But when we see other models, you will try to see how the other models work. So you blink every permission. So I like the blue going on through this, because there's a lot too much here actually. So you do, you can do it for yourself as a homework, you know, go ahead and do this homework. And I will probably take two classes for this for to see the flow chart.

And you can edit too, you can edit the flow chart, and you can create your own model, let's say their use linear regression right on to use by Bayesian linear system, and see how the data looks like. But you're supposed to know what you're doing linear regression is kind of Ready to wait, you know you are using the way but basically, you know, linear is a different manner. So you want to know what you're trying to do. So you want to know what exactly is the linear regression does what does elucidation treaters, you want to know this algorithms before you find this prediction? So mostly it was a prediction method. Okay, so we'll look into more two samples.

Okay. Two are three samples if you want, right, we can go through examples. If you think it's too less, no, but these things, so let's recap you know, let's recap. A distinction remains similar, clean the data as the data first thing is get the data, get the data without data, you cannot do anything, get the data and then you find out what to do with that data. If you see there is a None values the data is not good want to clean the machine data? Now this is important major part of that of the data, then you found out that that data is good, but you want it some specifics in the data, right?

Let's say you won't use some data, particularly data where there is no only information required for that particular information. So you want to use transformation logic will make SQL you want to use SQL Lite. use SQL, let's write SQL transformation. You want to use the SQL to get a particular amount of data and use only that data, right that manipulation you can do that here. After all these have been done, then you split the data and then use them I'm sorry, I didn't use the regarded up here use a diagram. for linear regression.

You can use any linear or base, you know, base linear. I'm sorry. That was basically there's no question it just any any other market is not doesn't have to be, we use only integration methods. The cluster method is a lot more methods a lot more reliable section, then you train it, you train it and get the score model. Once you get the score model, evaluate the model and see the results and the graph. Once you build a graph, visualize it, visualize the graph and see how does it looks like.

So this is what the histogram and then this is what the thing is showed up and you know, and of course, the data The predicted data, you cannot see the predicted the data. This is the score level that is the predicted data. So based on this history, it predicted the data how it will be? Okay. So we'll see this, why this was our experiment. This was a tutorial experiment.

We'll see two more or three if we have time. Okay. And and we will do more book. Yeah, notebooks and see if there is there is also an option for us to create tutorials. We have to do your info this here this is to do you see this? Probably.

We see this because I don't want to use any code because I'm not ready for that be using our code or C code or something. But let's see. Okay, let's see now what we'll do Here is how we add this to our project go to the project and scroll through here we go here and I want to add this Okay, here we go. Go to add assets. Now if you see here oops now we have so we have this income who don't have automobiles so we'll add this automobile and we train this modern automobile two, three of them will add it. Right.

It's good. We did some confirmation so so what happens here when you add this into the process, Jake? Um, no, that's cool. This again, we'll see here. So you have to you have this for, you know, two prediction, one transformation. And web services and for experiments for experiment for experiments typically actually do experiments, but one was for the web services to design that web services will create a predictive predictive experiment.

So that's the reason it comes on to actually what you want. Here you can see is, this is price prediction. This is predictive experiment. These incoming two, there's the reason when you're using this web services site, it creates one more bytes, and you can add more services if you want. Okay. So that's pretty much for our class.

Okay, we'll stop here for this experiments to an extent was actually one experiment but too much of actually there is a lot more more than this actually. But, you know, tomorrow for leads to mercury. So we go through another experiments let's see in this experiments we will see how it works and opening the studio and we'll see bail flow chart and try to execute that prediction and try to get results and we will modify that. Try to see the get the prediction that okay, that will help to understand more about our so so what is this all about? So this is you Machine Learning Studio is all about creating the flow chart, and predict getting prediction. You're training a model based on a data which is a Already, they're gonna have to create another item.

But you can create longer than this a different story. But that's a different workout together, we are not doing that. What we are trying to do here is using an algorithm, you're predicting the model. And using a score model, you're actually getting the data set here. And you put the evaluation model you are using the graph and getting the mean square root mean square and in all the information there to how the the model actually done. So if you're not satisfied with that model, you can change the split data, right?

So that's where spring data comes into picture. 1.7 5.5 point 8.1 I don't know. Okay, so it's, it's based on that one too. So it's also depends on that. Right. So I hope you found this useful.

I know this is This is a bit complex, because you will not be using for yourself it is more for machine learning. So but you can create this model, this is your Euclidean model, and you can deploy it into the machine so that it does automatically for you. Right? It's, it's not for you, but it's for machine. That's, that's the name like machine learning. So it's from a Microsoft Azure Machine Learning Studio.

That's what it is. It's what machine learning not for human learning for you. That's the reason you might not understand some of those things. Why this is generated. Why is what is this graph is all about why we are doing this. The whole point is to predict the model.

We are training the model and predicting the column data from the data. It's all about the data. You're, you're getting the data, you're deeming the data, and you what data you want, if you want SQL query for that SQL, and you're printing that particular weekend frame one column at a time, probably more. I will not use that. But But we use the one column editor. That would be simple for a flowchart.

And then you're training it using a lot of what is already there, and then you're seeing the results. That's what you're trying to do. Okay. And then, of course, you're generating the web services, which automatically moves the flow charts and, you know, makes use of prediction and generates the result. And you can view that to did I showed you that web services, so we can view that too. Let's not only are we going to experiment and see there will be this this is a prediction right?

So here a dot automatically change in which services import you see here you can view visualize this and see this this that is different from it it's not probably it may not be that different from that one. But you can correlate with that one with training experiments, see what holds the difference. Okay, so with this we in this tutorial will not take too much time here. Okay, we'll stop here. And tomorrow I get back to you or for me, it's tomorrow, Saturday for me the new model. We go to the I'm sorry, New Year of course that will be a new flow chart and see that Try to edit ourselves try to see or that will be in again an experiment.

So this is all the experiments we are we are working on experience, okay? And see how it works. And if put more transformation logic put SQL, I don't know join the data if you want to remove duplicate the cards, of course, you won't do that if you if there is there. So this is a data manipulation. So it's all about data. It's all about data.

You're doing machine learnings, dimensional data, predicting the future based on all the previous data. Okay, so that's all about so that's pretty much and I hope you may not understand everything in this video probably want to view this or probably want to view more demands are if you have any questions you can reach out to me. I love to see if I did see something because this is a new thing actually. It's pretty You started in 2018. And you have free access to make use of it, create your own murder, deploy it, you have option of creating your own, probably might invent your own model and do this. It's good.

No, it's good to know. You can do it your own thing. I mean, this you're not restricted with what we have done for price you might do for any other thing and publish it. You can publish to galleries here. You can predict your own thing, right. So So we'll start here, and enough.

Get I'll get back to you in the next class with a new model and new prediction. And meanwhile, thanks for watching, and I hope you like this video.