Hybrid Neural-Genetic Models for Adaptive Online Learning Resource Allocation

Abstract

The fast emergence of e-learning solutions has resulted in the rising popularity of smart systems that provide learners with resources adaptively. At present, when dealing with adaptive online learning solutions, there is no need to apply conventional resource allocation techniques because of the inefficiency of static scheduling procedures. Therefore, to eliminate the above shortcomings of conventional online learning resource allocation techniques, we suggest a novel hybrid solution called Hybrid Neural-Genetic Model for Adaptive Online Learning Resource Allocation that integrates artificial neural networks and a genetic algorithm for model optimization. Specifically, the artificial neural network is utilized for the analysis of learners' interactions and the prediction of their needs, while the genetic algorithm is applied to enhance the effectiveness of allocating resources and making decisions about it. The offered technique has been tested by using an extensive dataset concerning learners' interactions containing around 50,000 records. The experiment involved a number of assessment metrics, including accuracy, precision, recall, F1 score, efficiency of resource utilization, Mean Squared Error (MSE), adaptive allocation response time, and learner engagement score. The hybrid algorithm provided an adaptive allocation accuracy of 96.4%, which is 7.2% better compared to the accuracy achieved through the use of the neural network-based adaptive system and 10.8% more accurate than the one developed by the means of the genetic algorithm. Moreover, the introduced framework significantly reduced the prediction error of the system (MSE=0.028) while improving the efficiency of resource utilization to 94.7%. Finally, the proposed approach helped reduce the response time of the adaptive allocation from 250 ms to 176 ms while increasing the learner engagement score to 0.91. Thus, it is possible to state that the combination of predictive neural learning and evolutionary optimization can prove to be a valuable approach to implementing online education and may be easily scaled to the needs of bigger environments. The introduced hybrid approach seems to strike a balance between user requirements and computational costs while providing a solution capable of making decisions in real-time. This research is useful in terms of developing intelligent systems for education.