College Predictor Using Machine Learning

Introduction (Problem Statement)

Being a first-year undergrad at IIT Bombay and having gone through the whole month-long counseling process after taking the entrance exams, I am well-versed with all the problems that an average engineering aspirant in India faces. The main problem during the counseling process arises when you have to manually look at and search the colleges you are eligible to get admission into based on your rank in the Joint Entrance Exam. If somehow we could get to know a basic idea beforehand, all the hassle would somehow get reduced. With the same thing in mind, I created a rank predictor using Microsoft Azure Machine Learning Workspace.

All that the rank predictor does is, take in the "closing rank" data of the previous 10-12 years, analyze it, apply an algorithm, and create a specific model through which it, if not deterministic but probabilistic, generates the closing rank for that specific college for the year that one is concerned about.

Creating Azure ML Workspace

1. Go to portal.azure.com
2. Go and search "Azure Machine Learning Workspace"
3. Fill in all the details, and there you have your own workspace created right before your own eyes.

Feeding Data

1. Launch your Azure ML Workspace studio.
2. Go to "Data" under the "Assets" column.
3. Click on "create" a data asset and fill in all the required details according to the data that you are going to train your model on.

Developing a pipeline

Azure pipeline is a workplace where with the help of Graphical user Interface, we will be training our model with the data that we have.

1. Go to the "Designer" tab.
2. Click on "Create a new pipeline".

The pipeline should look something like this:

Click on "configure and submit" and wait till the process of model training gets over.

Creating a real-time inference pipeline

So the real-time inference pipeline is the place where you can use your model to feed known data either manually or by some other web input and then the model should help you predict the unkown data using the training that it had undergone before.

The real-time inference pipeline should look something like this:

Click on "configure and submit" and wait for the whole process to get completed.

Getting the output

Click on "score model", and your predicted rank should be visible under the "Scored Labels" column.

Attaching the excel sheet with this article containing the raw data as well as the predicted data.

Note: I have made a video on the same. The video will make up for a great reference. 

P.S.: A big thanks to Mahesh Chand, the CEO of CSharp Inc., for motivating me to dive deep into the tech world. I remember meeting him at a tech conference in India and was greatly influenced by his knowledge of technology.