Predicting Student Performance Using Deep Learning and Indian Knowledge Systems | Chapter 8 | New Horizons of Science, Technology and Culture Vol. 8

 

Background: Existing data-driven approaches have demonstrated promising predictive capabilities; however, many remain narrowly focused on technical optimisation. By treating behavioural data as context-free signals, these systems often overlook the cultural, social, and ethical dimensions that influence learning.

 

Aim: The aim of this chapter is to develop a culturally grounded and ethically responsible framework for predicting student academic performance by integrating mobile phone behavioural analytics with principles drawn from the Indian Knowledge System (IKS).

 

Objectives: The study seeks to (i) model student learning behaviour using temporally rich mobile usage data, (ii) enhance prediction accuracy and interpretability through deep learning architectures, (iii) operationalise IKS-inspired constructs to provide cultural and ethical context, and (iv) support fair, human-centred educational interventions.

 

Methods: An integrative multi-input deep learning framework is proposed that combines Temporal Convolutional Networks (TCN) for sequential behaviour modelling, attention mechanisms for feature prioritisation, and static psychometric and demographic feature fusion. Mobile phone data capturing usage patterns, activity rhythms, and engagement indicators are processed alongside IKS-informed contextual features. Model performance is evaluated against classical machine learning baselines using predictive and fairness-aware metrics, with interpretability analyses supporting transparent decision-making.

 

Results: Experimental validation on representative datasets demonstrates that the proposed framework consistently outperforms traditional machine learning models in terms of prediction accuracy and stability. Attention-based explanations reveal that IKS-inspired features contribute meaningfully to performance gains while reducing subgroup disparities. The results indicate improved fairness, enhanced interpretability for educators, and greater alignment with student well-being and learning rhythms.

 

Conclusion: The results indicate that embedding contextual and value-oriented dimensions improves not only predictive accuracy but also interpretability and fairness, making the system more aligned with real educational environments. The inclusion of explainability mechanisms further strengthens trust and transparency, which are critical for adoption in academic settings. Overall, this work contributes a meaningful step toward educational AI systems that support holistic learning, respect learner identity, and encourage responsible decision-making in data-driven education.

 

Novelty: The key novelty of this work lies in embedding indigenous knowledge principles directly into the design and interpretation of deep learning models for educational analytics. By bridging behavioural data, advanced neural architectures, and culturally rooted context, the framework advances a human-centric paradigm for academic performance prediction that emphasises ethical responsibility, cultural resonance, and holistic educational outcomes.

 

 

Author(s) Details

S. Vimala
Department of Computer Science, St. Joseph’s College (Autonomous), Tiruchirappalli -2, Affiliated to Bharathidasan University, Tamil Nadu, India.

 

G. Arockia Sahaya Sheela
Department of Computer Science, St. Joseph’s College (Autonomous), Tiruchirappalli -2, Affiliated to Bharathidasan University, Tamil Nadu, India.

 

Please see the book here :- https://doi.org/10.9734/bpi/nhstc/v8/7176