Predicting who will drop out of nursing courses: a machine learning exercise

Area of Science:

• Educational research
• Data science in education

Background:

• Distinguishes between causation and prediction in academic practice.
• Highlights limitations of causation-focused studies in nursing student attrition.
• Proposes prediction as a more fruitful approach to understanding attrition.

Purpose of the Study:

• To apply rule induction for predicting nursing student attrition.
• To evaluate the effectiveness of the Answer Tree package for this purpose.

Main Methods:

• Utilized rule induction via the SPSS Answer Tree package.
• Employed a dataset of 3978 records from 528 nursing students.
• Split data into training and testing sets for model validation.

Main Results:

• Achieved 84% sensitivity in predicting attrition.
• Reported 70% specificity for the prediction model.
• Demonstrated 94% overall accuracy on unseen data.

Conclusions:

• Rule induction is effective for reducing student attrition when high-quality data is available.
• Suggests comparing algorithmic predictions with tutor-based assessments.
• Emphasizes the need for substantial, reliable data for successful implementation.