Students Enrollment Projection

Students projection service for university enrollment using machine learning forecasting.

Abstract

Analyzed over 10,000 database rows to implement and deploy student projection service for university enrollment by designing a web graphical interface and statistical backend service. Awarded as the best project.

Statistical model

Our model consists in a linear regression model that predicts the number of students that will be enrolled in a given semester and course.

Using the data from the university enrollment database of previous years and Python libraries. We analyzed different variables with the help of heat maps, linear plots and so on.

We found that the most important variables are: average grade, number of students enrolled in the previous semester and average number of times the course is taken.

Architecture

Our system is composed of 5 main parts: frontend, backend, statistical endpoint, database and university enrollment database which is read only.

Each of these parts is implemented in different languages and hosted on different servers.

The whole system is protected by an OAuth2 authentication system which provides a token to access the backend and statistical services. The token is valid for 1 hour.