Unlocking Literacy Success: Leveraging Machine Learning to Identify At-Risk Students

• In Event: AERA Roundtable Session Thursday 9:50 am Four Seasons Ballroom 4
  In Roundtable Session: Technology as an Intervention of Learning (Table 3)

Abstract

As literacy skills are a fundamental aspect of early childhood education, being able to identify students who will have difficulty can be beneficial for students and educators. This study leveraged machine learning to develop predictive models to determine which pre-K students will continue to struggle with learning early literacy skills during a kindergarten readiness program. A large dataset of pre-K students was utilized to develop several decision tree models, such as C&R Tree, C5, CHAID, and Exhaustive CHAID. The models were compared across performance metrics, and results indicated that the Exhaustive CHAID models were especially well suited to predict struggling students with relatively high accuracy and recall.

Authors

• Haya Shamir, Waterford Institute

• Mikayla Tom, Waterford Institute

• Erik H. Yoder, Waterford Institute

• David Pocklington, Waterford Institute

• Alexander Houchins, Waterford.org