Admit Predictor

Predicts the probability of a student to get into an university of their choice. Helps students save on those expensive application fees.

Comment

Inspiration

There are hundreds of colleges for a student to apply to when pursuing higher education, a student may for example have 5 colleges on his/her mind to apply to but has no way of knowing if he/she actually has chance to get into those universities. The average application fee most colleges ask is around $60; hence that student to apply to those 5 colleges has to spend $300 and if rejected by even 3 of those ceases to lose $180 and also the time spent of posting the application and writing other letters for it. Being final year students, we wanted to try pursuing an MS degree in the USA after our graduation and we faced the same situation mentioned above and got an inspiration to do with project with the goal of helping fellow students.

What it does

The purpose of this project is to develop an ML model which predicts the probability of a student getting admitted to the college of their choice based on their profile (which includes GRE score, IELTS score etc).

How we built it

We built it React on the front end with material UI, testing with JEST and Enzyme. CI/CD was implemented with GitHub Actions and deployed on GitHub Pages. The backend was built on Django and Django Rest framework with the ML model being built with Tensorflow with an accuracy of 97%. CI/CD for the backend was also done with GitHub Actions and deployed on AWS Lambda through AWS API Gateway with AWS Elastic File Storage. CodeCov was used for measuring code coverage.

Challenges we ran into

As we wanted to follow a methodology of Serverless Compute we decided to use AWS Lambda for our project. One of the challenge we faced was deploying our project which was based in Django to AWS Lambda, after research we found we had to use Zappa for the same and we did. As soon as did challenge was we faced another one which was AWS Lambda had a file upload limit of 250Mb but our project zip file was very well above it because of Tensorflow module. Hence to overcome this we again after a lot of searching we found out about AWS Elastic File Storage and how we had to mount it on EC2 and connect it to Lambda.

Accomplishments that we're proud of

"An investment in knowledge always pays best interest".
We accomplished a lot through this project, but the one we are most proud of is the knowledge we gained throughout this experience.

What we learned

We learnt many new things especially in the cloud as such as using different services of AWSs, Ml models etc.

What's next for Admit Predictor

For now Sop and Lor was ratings were in the data set was dummy values but in future we plan to use sentiment analysis and ask the user to upload their letters we so can use NLP and rate them.

Built With

Share this project:

Updates