Exploring the Machine Learning Paradigm in Determining Risk for Reading Disability.

Area of Science:

• Education
• Psychology
• Computer Science

Background:

• Reading disabilities (RD) pose significant challenges to academic achievement.
• Accurate risk determination is crucial for timely intervention.
• Existing models may face limitations with complex data relationships.

Purpose of the Study:

• To evaluate the predictive performance of machine learning (random forest, RF) against traditional logistic regression (LR) for determining reading disability risk.
• To assess the effectiveness of a constellation model incorporating multiple RD indicators.
• To compare the utility of RF and LR within the constellation model framework.

Main Methods:

• A constellation model was used to operationalize third-grade RD risk for 12,171 Florida students.
• RD risk was determined using one to four RD indicators from first and second grade.
• The prediction accuracy of RF was compared against logistic regression (LR).

Main Results:

• Both logistic regression (LR) and random forest (RF) demonstrated comparable accuracy in predicting RD risk.
• Reading fluency emerged as the most significant predictor of RD risk across models.
• RF did not outperform LR when dealing with multicollinearity among predictors.

Conclusions:

• Random forest (RF) offers no significant advantage over logistic regression (LR) for RD risk prediction in models with linearly related predictors.
• Reading fluency is a critical component for early identification batteries aimed at determining later reading disability risk.
• The constellation model effectively integrates multiple indicators for a comprehensive assessment of RD risk.