How does Shiksha Data Science model calculate the chances of admission?

This article describes Shiksha Study Abroad’s Data Science model that predicts the chances of admission of a student in a particular course of a university.

At ShikshaStudyabroad, we aim to provide students access to transparent information and data-based tools that can help them make informed decisions in their study abroad journey. This model brings together our expertise in the field of university listings and counseling knowledge under a single umbrella.

Why know your chances of admission?

Most people will agree that deciding the right university to study is the most important and complex decision that a student has to undertake. It’s complex because unlike Indian universities, which have a relatively simpler exam cutoff based admissions system, universities abroad look at student profile more holistically by giving weightage to both objective factors (such as exams score needed to study abroad, academic percentage) and subjective criteria (such as quality of work experience, professional recommendations, and career goals). Not only do these weightages vary by universities, but it is common to see different courses/departments within the same university giving different weightage to these parameters. However, very few colleges will clearly lay out the exact assessment criteria, offering loose guidelines at best. Such an environment creates an information asymmetry for international applicants. Knowing one’s chances of admission can therefore remove this asymmetry and help students make better decisions.

Why do we need a Data Science model?

Currently, all requests for a ‘Rate my chance’ on a particular course and university are sent to an expert counselor, who reviews the student profile on objective parameters and then shares a manual rating. Although these ratings are quite accurate, (as Shiksha SA has experienced counselors who have worked with hundreds of students and are trained by the university representatives) but each rating requires manual intervention. This is an expensive and time-consuming process, so much so that Shiksha SA has to limit the number of requests each student is allowed to undertake to avoid misuse. The new Data Science model introduced by Shiksha SA will provide ratings in real-time while trying to maintain a high accuracy of prediction.

How does the Shiksha Data Science model work?

The input used for the Data Science model is the objective criteria of a student profile (e.g. academic percentage, backlogs, work experience, exam score, etc.). And the output is a probability score of admission and this is further classified on a 5 rating scale:

1.     Poor (less than 5% chance of admission)
2.     Very low
3.     Average
4.     Good
5.     Excellent (greater than 95% chance of admission)

The accuracy of any good Data Science model relies on the quality and quantity of data used to create it. This is where we run into complexity in creating a model for such a requirement. To create such a model, hundreds or thousands of diverse student profiles of international applicants for each course and university are needed. The problem is many of these courses and universities do not have large enough applicant pools of international students and also there are no known sources that publish this information. To overcome this limitation of sparse or missing data, we have substantiated student profile data along with counselor expertise so that the model has sufficient accuracy and can be applied to a diverse range of profiles.

What are the limitations of this model?

As mentioned earlier, foreign universities look at both objective factors (such as exam score, academic percentage) and subjective criteria (such as quality of work experience, professional recommendations, career goals) before an admit or reject. The data science model only works on objective parameters and cannot take into account any of the subjective criteria. Sadly our current machine learning & artificial intelligence systems haven’t reached that level of sophistication to fully remove the role of a human counselor as yet :-) For a more well-rounded assessment, we still recommend you speak to an expert Shiksha counselor.

Currently, the data science model only works for a limited set of course and country options. As you use this tool, please share your feedback so that we can make improvements in its future versions. Please drop me an email or post it as a comment on this article below.