Predicting and Analyzing Academic Performance of CITU CCS Students Using Hierarchical Neural Networks and Augmented Clustering: A 16PF and CFIT Assessment Approach

A hierarchical neural network was used to predict scores in both Intermediate Programming and C Certification using cognitive and personality assessments. The model, which integrates the Cattell Culture Fair Intelligence Test (CFIT) and the 16 Personality Factors (16PF), achieved a mean accuracy of 95.15% (±1.97%) and a mean loss of 0.2629 across 10-fold cross-validation, significantly outperforming traditional machine learning models that achieved accuracies of only 47-59%, while also assuring its ability to generalize. Beyond predictive modeling, an augmented clustering approach was employed to profile students into performance-based clusters, supported by statistical significance testing through t-tests across high performing and low performing groups. Results revealed key differences between high- and low-performing clusters, such as top-performing students exhibited higher cognitive ability (mean CFIT score = 9.33, p < 0.01), logical reasoning (attr_B = 6.52, p < 0.05), and emotional stability (attr_C = 5.63, p < 0.05), along with lower dominance (attr_E = 3.15, p < 0.05) and impulsiveness (attr_F = 4.15, p < 0.05). This is in comparison to low-performing students, which demonstrated significantly lower cognitive scores (mean CFIT score = 3.00, p < 0.01) and higher dominance (attr_E = 5.62, p < 0.05) and impulsiveness (attr_F = 6.29, p < 0.05). Despite the promising results, it is importance to assert the effectivity of such a model across interdisciplinary fields.