A Novel Deep Learning Technique for Big Data Anomaly Threat Severity Prediction in ELearning

Abstract

E-learning platforms are susceptible to several anomalies, including abnormal learning behaviours, system abuse, and cyber security (CS) attacks, which interfere with the learning process. Conventional methods for detecting anomalies have limitations with high-dimensional data, skewed distributions, and poor feature selection, resulting in incorrect severity level predictions. To overcome these, a novel Sea Lion Multilayer Perceptron (SLMP) model is introduced for anomaly severity level prediction. At First, an e-learning anomaly dataset is gathered and trained in a Python environment. Hence, the data is preprocessed, and the Sea lion optimization (SLO) is used to select the best features to attain only the most significant attributes. Subsequently, the chosen informative features are employed for further process. Moreover, prediction and classification are performed using the SLMP model. Finally, Performance metrics like F score, Accuracy, recall precision and error rate are used to evaluate the effectiveness of the model. The results confirm the efficacy of the developed SLMP framework over current methods, illustrating its strength in optimizing predictive efficiency for anomaly severity detection in e-learning systems.