msAzureVideo

well a bit of introduction about myself yeah this is Kuljot first year

undergrad at IIT Bombay at the Department of Environmental Science and Engineering batch of 2026 yeah so we'll be looking at rank prediction using MS

Azure ML workspace so the first question is what does the rank predictor do so

basically I've created a rank predictor model using machine learning from Microsoft Azure portal and the rank predictor uses linear regression algorithm which is a

mathematical algorithm which we are gonna dwell deeper in the latest slides and so we are using

this mathematical algorithm to predict the closing ranks of various engineering discipline branches

at various IITs by observing the previous 10 to 12 year trends so what my model does is I feed

data into the model that is the closing ranks of various branches at various iits starting right

from the year 2008 to the year 2023 then the model uses linear regression mathematical algorithm to

observe the pattern the various pattern changes the various trends starting right from the year

2008 to 2023 and then it creates a sort of linear equation which again we are gonna look into in the

later slides and then using that linear equation the model predicts the closing rank for the year

2024 or for the for whatever year we're concerned about and for this model I've used MS Azure

machine learning works workspace to to predict the ranks so talking about the algorithm our

algorithm is linear regression and this algorithm is supposedly the most basic algorithm that

could be used for a model like the one that i've created so so basically there are two types of

linear regression algorithm one is the multivariate algorithm and the second is univariate algorithm

and as the name suggests the difference between the two is very basic so uh in multivariate in

multivariate algorithm you have one uh dependent variable and one independent and sorry and

multiple independent variables whereas in univariate linear regression algorithm you have one uh

dependent variable and one independent variable and the model that i have created uses

univariate linear regression where wherein the dependent variable would be the closing

rank and the independent variable would be the year that we are concerned about

now in case of a univariate linear regression the equation is you can correlate the equation to that of a straight line that represents the relationship between the dependent and independent variables and

Moving on to the next slide Okay, so basically What I have here is a graph of univariate linear regression

Wherein y this y over here is the dependent variable and the x over here is the independent

variable so from this you can see the dots initially that is the points that are plotted

on the graph initially are not in a straight line but what we are doing with linear regression is

where we are creating the the equation of a straight line and we are going to use the

equation of this straight line over here to basically predict to basically plot y for a

specific value of x so in my model the y over here which is the dependent variable is the

closing rank of the branch that we're concerned about and the x that is the independent variable

over here is the year that we'll be talking about so the equation looks something like and sort of a

straight line y is equal to a plus bx where y is again the dependent variable x is the independent

variable a is the intercept on the y-axis so what I mean by intercept is if I produce the straight

line backward the point at which this straight line is going to cut the y-axis would be my

intercept that is the value of a and b is the coefficient of x which can be easily found out by

writing the values of y for the respective values of x and by elimination method we can easily find

the coefficient of x so you can correlate this coefficient of x by i mean by the slope of a

right so yeah that's pretty much it for this slide I guess yeah moving on to the next slide

okay now the main question here is how reliable is my model so the data is predicted or collected

using mathematics that is linear regression algorithm is used and obviously mathematics

doesn't like but the main point here is the sentiments of the people or the

industry trends aren't taken into account so I mean these industry trends

or the the thought process of the people can obviously be tracked and monitored

using AI enable tools which again is an advanced case so but for my model I

I haven't done it. So, uh, if any, uh, uh, I mean, if, if

if any random variation is observed, uh, please forgive me, I guess

And yeah, that's pretty much it. Okay. So for the model that I here is uh the mean absolute error of the model would be 77 That means the rank let say the rank that i predicted was 200 around 200 for uh let say iat root key computer science so that means the

the rank can vary between 200 plus or minus 77 yeah and the coefficient of determination here is

0.998712 which is approximately less than 1 by a magnitude of 0.01 and so the principle here is

as close the coefficient or determination is to 1 the more precise or the more accurate our model

works in mathematical sense so obviously my model in mathematical sense is working quite great and

yeah that's pretty much it about this slide. Okay so the excel sheet that I'll be providing will

look something like this. So what you have here is the closing and closing ranks for

various branches from year 2008 to year 2023. So this is so the data here is already available

on various websites and the data for the year 2024 is the predicted closing lines for the various

branches and I mean for to be on the safer side I'll get the error as 200 plus or minus 200 that

means let's say if the prediction closing if the predicted closing line for IIT Bombay Aerospace

department is 3.115 that means the actual closing rank can range from 3.115

minus 200 to 3.115 plus 200. Okay. So that's it for the PowerPoint presentation

Yep. So now we'll be moving on to the Azure ML workspace portal

So let's go to designer. Yeah. Final line prediction

So yeah, the designer is basically the place where I train my machine learning model

And so it doesn't require much of coding. It's just a basic graphical user interface

So here is the algorithm that I've used, the dataset that I used for training This is the train model area where it uses the linear regression algorithm and trains the model by using linear regression algorithm on the defined rank dataset that has been provided to it

So there's the score model and the evaluate model function here. So to show things I have..

Yeah, so that's the data that I provided in terms of comma separated value. Yeah

Now moving on to basically configure and submit

Go to view details

So after training the model, the score model looks something like this

Yeah, so this is the predicted column. So I've trained the model on predicting year 2023

And so this is the predicted column. And this is the data that I have already provided

So basically the model checks whether the value, this value is close to this value or not

and then accordingly it provides me the error value and I mean group mean

squared error value and stuff like that okay so that is the train model area now

moving on to the real time model area yep

So, in this I have provided the data manually to the model

Now this is the SPOR model area where the model has predicted some of the ranks for the year

that I am concerned about that is the year 2024. So this is the predicted rank area and this is

wherein I have provided the values in terms of comma separated values to the model

so yeah that's pretty much it I guess hope you enjoyed it

thank you