Technikon Pretoria uses business intelligence to enhance first-year student retention

Technikon Pretoria is using leading business intelligence software to develop a model that ensures improved first-year student retention.

"Tertiary institutions worldwide face the problem of students not completing their first year successfully, and taking too long to graduate," says Izak Smit, statistical analyst at Technikon Pretoria's Department of Statistical Support.

Smit and Dr Amanda Lourens, director research and development at Technikon Pretoria, recently completed an extensive research project on student retention at the Technikon.

Smit added: "We wanted to develop a statistical model to predict first year success. "Using SAS business intelligence software, we are designing a model that can assist in improving first year student retention, get students to finish more quickly and limit the number of dropouts.

"SAS has always been considered the Rolls Royce of statistical software as it handles large data volumes and has huge manipulative power," says Smit. "In addition, there is a lot of information available around SAS and we found literature very appropriate to our particular study."

The Department of Statistical Support has used SAS software for more than five years, and has recently renewed its licence. The department provides free statistical support to post graduates and staff, as well as external commercial consultation.

"For our students and staff, we help with experimental design, sampling protocols and questionnaire design," says Smit. "Once data has been collected, we help with the analysis, providing feedback and interpretive assistance."

One of the most exciting current projects is the retention study.

"We wanted to use our operational database to predict the probability of students being successful in their first year," says Smit. "We have 5 000 student records and we classified each as a successful or unsuccessful first year student. We then used 12 variables, including gender, matric aggregate, race, age and English marks, to see if any of these actually predicted first year success."

Using stepwise regression, eight variables which were the best predictors of success were retained in the model. A section of the 5 000 records was then used to build and calibrate a statistical model.

The remainder of the data set - unused in the model calibration - was then utilised to validate the model.

"We were thrilled to find that our model predicted first year student success accurately in 75% of the cases in the validation data set," says Smit.

The technikon will now extend this model, using additional variables including non-academic factors.

"Our follow-up study will include factors such as social conditions and study environment, for example whether students also have part-time employment, and how far away they live," says Smit. "This will further improve predictive accuracy."

The model, currently generic, will also be taken down to faculty level.

Smit foresees the model being used by the technikon for three main purposes. Firstly, it can help with recruitment by describing the type of person who would probably be a successful student.

It can also indicate which course would be best for a particular person.

"We can predict that a person is unlikely to be successful for one particular course, but likely to succeed in another," says Smit.

Thirdly, the model can identify students at risk of being unsuccessful. They could then be given priority, and placed on a foundation or bridging course.