Enhancing Gifted Identification: A Machine Learning Analysis of the HOPE Teacher Rating Scale Responses

Abstract

This study explores machine learning (ML) approaches for identifying gifted students by integrating academic and socioemotional characteristics from the data collected with the Having Opportunities Promotes Excellence teacher rating scale. By using the Gaussian Mixture Model (GMM) and ML approaches, including support vector machine (SVM) and multilayer perceptron (MLP), the study analyzed data from 1,157 elementary students in South Korea. Results showed strong correlations between academic and socioemotional score factors (r = 0.8), with weighted GMM effectively clustering gifted students while addressing data imbalances. Classification models revealed high accuracy but highlighted challenges in terms of identifying underrepresented students. In addition, Bias analysis indicated significant variations among teachers in rating scores across socioeconomic and ethnic groups, which may have implicit biases in evaluation practices. The results underscore the importance of evaluating current identification practices that account for multiple domains of giftedness as well as mitigating teacher biases.