A person-centered approach to cognitive performance analysis in primary school children: Comparisons through self-organizing maps

Abstract

The objective of this study was to identify distinct student profiles based on physical, psychological, and social characteristics, and examine their impact on cognitive performance. A total of 194 children participated in this cross-sectional design study (mean age = 10.61 years, SD = 0.45; 48.96% girls). The study included participants from diverse racial backgrounds. Using Self-Organizing Maps, an unsupervised neural network clustering technique, six distinct profiles were identified. These profiles revealed significant effects in daily physical activity, self-reported physical, social, and psychological factors, and physical performance. Profiles characterized by higher physical activity levels and positive social and psychological factors were associated with better cognitive performance, in contrast to profiles with lower levels in these domains. These findings suggest that students' cognitive outcomes may be linked to their physical, psychological, and social characteristics, which interact to shape cognitive functioning. The recognition of the diversity of student profiles in specific educational settings may facilitate the design of more targeted programs that address individual needs and strengths, thereby enhancing their development in these domains within similar educational contexts.

Fig 1. Flowchart for the unsupervised SOM algorithm modified from Seifert et al. [56] with permission.

Here, k denotes time, x(k) is an input vector randomly drawn form the input data set at time k, h_j,i denotes the neighborhood function around the winner unit j and η(k) the learning rate at time k.

Fig 2. Description of results of self-organizing map analysis.

The component planes from the analysis are shown in panels (A–C) for the physical, psychological, and social variables of the three domains, respectively. Panel (D) displays the Davies–Bouldin index related to the k-means cluster algorithm applied to set profiles. Panel (E) shows the obtained clusters and the hits (neuron shadows). Panel (F) shows the number of participants in each cluster, the girl/boy percentage in each group and the results of the chi-square test.

Fig 3. Pairwise comparisons of cognitive domain variables between clusters.

The box plots display median values, interquartile ranges, and outliers for each cluster across three variables: Academic achievement (left), perceived mathematical performance (center), and perceived language performance (right). Significant differences between clusters are indicated by asterisks (*p <  0.05).