An Intelligent Learning Analytics Framework for Predicting AI

Proficiency Using BER-SC-Optimized MLP

Hessa Al-Junaid1,*

1 Department of Computer Engineering, University of Bahrain, Zallaq, Bahrain

Received: August 08, 2025 Revised: October 16, 2025 Accepted: December 11, 2025 ⋆ Corresponding author

ABSTRACT

The fast adoption of artificial intelligence (AI) and machine learning (ML) in the sphere of higher education has

become a catalyst that requires more precise and data-driven models that could assess the level of proficiency and

interest in new technologies in students. This work is driven by the fact that there are currently no personalized

and analytically rigorous frameworks for assessing the level of AI literacy among college students and proposes an

optimized learning analytics model that hyperparameters a Multilayer Perceptron (MLP) by utilizing the Binary

Al-Biruni Earth Radius with Sine Cosine Algorithm (BER-SC) metaheuristic algorithm to close the performance gap

between human and artificial intelligence. The suggested BER-SC + MLP hybrid was created and evaluated using a

sample consisting of 258 student responses from Grand Canyon University, which encompassed AI knowledge and

usage behavior as well as career interest. The model achieved an accuracy of 0.9312 after optimization, compared to

the MLP’s pre-optimization accuracy (0.8996), and a significant decrease in Mean Squared Error (MSE = 0.0002808).

The results demonstrate that BER-SC is far superior to nontraditional optimization techniques in terms of convergence

efficiency and prediction accuracy. The contribution of the study is that it created a scalable and interpretable, and

high-performing educational data modeling framework that has the potential to support adaptive educational learning

analytics and institutional decision-making. The results suggest that intelligent optimization added to the ML-based

assessment systems has a high possibility of enhancing the quality, efficiency, and feasibility of AI-based education

and allowing more adaptable, more affordable, and more individually focused learning settings.

Keywords: Artificial Intelligence (AI) in Education Machine Learning (ML) Optimization Multilayer Perceptron (MLP)

Binary Al-Biruni Earth Radius with Sine Cosine Algorithm (BER-SC) Educational Data Mining (EDM)

1. INTRODUCTION

Artificial intelligence (AI) and machine learning (ML) have

become disruptive technologies over the past few years and

can affect various areas, such as healthcare and finance, engineering

and education. Nevertheless, the lack of proper

and systematic approaches to the development of AIrelated

skills in the context of college students is one of the most

significant issues that dominate the educational discussions

today [1]. Although AI and ML technologies are becoming

more prevalent in the professional fields, universities tend to

provide a limited number of students with the background

and practical skills necessary to implement such technologies

effectively [2]. There is an increasing difference between

knowledge of AI and its practical application, which creates

significant concerns in how universities can train students to

become data-driven and technology-intensive workers. The

origins of this challenge can be traced to a number of factors,

which are interdependent. First, the abstract conceptual nature

of AI is a natural challenge to the student especially when