The idea of predicting the future – whether a short or long period of time – is very appealing. Did you know that it’s already possible? And not just that: it can be very helpful in many fields. From Finance to Health Care, and Higher Education in particular. See why embarking on a predictive analytics journey may lead you to attain better results, from great student experiences to financial benefits.
- The first question to consider is how student success is defined within your institution. This will help you set the targets for a prediction. For example, which characteristics about a particular student do you want to predict and how you plan to use that information.
- Identifying at-risk students, a concern which Predictive Learning Analytics is frequently used to resolve.
- Building student retention or grading models based on demographic factors, with the goal of identifying groups of students who are not being assisted as well as they could be.
- Developing student success models in relation to: employment after graduation, engagement levels, or the ability to be self-driven and self-regulating about their own lifelong learning.
- Predicting learning processes and engagement.
- System usage to identify students that can improve from average to excellent, with a little assistance.
- Information about students’ abilities or prior achievements (SAT scores, high school GPA etc.).
- Details about their activity (course attendance, LMS usage, grades on submitted assignments, among others).
- Features about their learning environment (major, class size etc.).
- Demographic data such as gender, race, family background, among others, is another possible source of predictors, but one that must be used with caution not to embed historical bias into the model.
- Once this information is gathered, a model can be built to represent relationships between the predictors and the prediction target for this known data, and then applied to incoming students. The known input data can be used to make a prediction about the likelihood of a student dropping out or earning a low grade, for example.
- Increase in student success. More specifically, the number of students who are likely to complete their degrees.
- Financial benefit for both institutions and students, since there is a tendency in lowering the number of college drop-outs (less impact on tuition payments for institutions and a higher chance of a financially successful career for students). In fact, investing in systems and processes can have an immediate financial return.
- Improved learning experience.
- Achieving student excellence more often. Student apps are available to provide them with information about their engagement level with their studies and to help them select courses in which they are predicted to succeed.
- Predictive analytics should always be accompanied with strong follow-up. For example, if a student was flagged with high academic risk and an intervention was conducted, the results must be examined closely by the educational institution.
- Putting a feedback mechanism in place that adjusts and improves the predictive algorithm over time is also important. Otherwise, the risk of perpetuating existing biases and inefficiencies can occur.
Alyssa Friend Wise, associate professor of Learning Sciences & Educational Technology at the Educational Communication and Technology Program of the New York University.
Dragan Gasevic, professor at the Moray House School of Education & School of Informatics at University of Edinburgh.
Niall Sclater, learning technology consultant.
Xavier Ochoa, professor of the Electric and Computing Engineering Faculty and director of the Information Technology Center at Escuela Superior Politécnica del Litoral, in Ecuador.