Predicting Learning Retention Using Hybrid Models of Spiking Neural Networks (SNNS) And Deep Learning

Abstract

This paper seeks to explore the issue of learning retention prediction which is very essential in any educational institution yet the problem of accurate prediction has not been adequately addressed by the existing algorithms because of the difficulties involved in dealing with temporality and feature interaction. To overcome these limitations, the proposed algorithm will employ a novel hybrid approach, which will integrate the capabilities of Spiking Neural Network (SNN) with deep learning approaches specifically CNN to predict learning retention. Proposed model name as SNN-CNN-LRN. By utilizing the strengths of both short-term and long-term behaviors, this approach will enable the model to predict accurately. Three main data sets have been used in this work namely the HESP dataset, which consists of demographic data with 31 attributes and 145 instances, the XAPI dataset consisting of learning experiences with 16 attributes and 480 instances, and lastly the HEI dataset comprising economic and institutional data with 36 attributes and 4424 instances. It is clear that the proposed model is capable of performing significantly better than other models, like CNN-BiGRU and RNN-LSTM, with an accuracy rate of 98.45%, precision rate of 98.25%, recall rate of 98.85%, and F1 score of 98.95%. This result clearly reflects the superiority of the model, and it can be used for developing adaptive learning methods.