Algorithm to Predict Student Performance

Project Description

Summary of business:

We are a private Christian university, serving both traditional on-campus students and non-traditional adult online students. 

Problem:

Adult online classes are five weeks long. This is a short window in which to identify students who are struggling in order to intervene. When a student fails a class, there are financial repercussions, they are less likely to return for another class, and they are less likely to graduate. Identifying these students as early as possible will better serve the students and the university.

Our team has received some formal training on data science, R, and predictive modeling. However, we do not have any real-world experience building predictive models and are seeking someone to act as a guide and walk alongside us during this project. We can do the bulk of the data preparation in-house, with guidance from the expert. The actual model building will be led by the expert with our input.

Expertise:

We are seeking someone who has experience building predictive models and understands the unique challenges of adult online higher education. The ideal candidate will have experience with R and T-SQL. 

Technology stack:

MSSQL, Wherescape RED (data warehouse automation/ETL), Tableau, R 

Data Sources:

Our MSSQL Enterprise Data Warehouse (star schema) contains a great deal of information about each student. This includes general demographics as well as their academic history.

We also have data from our online learning platform including individual assignment grades, student login information, and student behavior within the learning environment. The example file attached provides a rough idea of the types of data we plan to use.

Deliverable:

We are seeking to build an algorithm in R that can run daily and predict the likelihood of eventual failure for each enrolled online student. This algorithm will assign a score to every student/course/day and will write into a MSSQL database.

Location Preference:

We prefer US-based candidates but will consider international candidates who are willing to work during US business hours.