Akademik Veri Yönetim Sistemi
Egyptian Informatics Journal, cilt.32, 2025 (SCI-Expanded, Scopus)
Online Social Network spam detection has gained importance as malicious content abuses platform purity and user experience. Especially for the X platform, spam detection requires more effort due to its character limitations, multilingual content, and diverse symbol usage. The inherent uncertainty in spam characteristics and the dynamic properties of X content require advanced computational approaches with proper modeling of uncertainties while keeping up high accuracy under complex cybersecurity threats. This study presents a Self-Adaptive Interval Type-3 Fuzzy Logic System (SA-IT3FLS) with new hybrid learning approaches, integrating Square Root Cubature Kalman Filter (SCKF) and Extended Kalman Filter (EKF), to enhance spam detection accuracy for X platform against high levels of uncertainties. The proposed approach specifically targets the modeling of uncertainty challenges that appear because of the boundary between spam and non-spam. A type-3 fuzzy system is suitable for managing uncertainty at the membership function level, rule evaluation level, and within the data itself. The new adaptive approach is implemented to tune the rule weights and the membership functions (MFs) centers in this study. Along with the optimized covariance matrices, the initial values of MF centers and rule weights are optimized with Particle Swarm Optimization. The optimized parameters are learned by SCKF and EKF. The proposed method's effectiveness is revealed through four scenarios with real-world data sets. The experimental results of the proposed method showed robust performance, achieving an accuracy of (96.74%),a recall of 0.98, an F-score of 0.96, and an AUC value of 0.98, showing a conspicuous increase in the performance. Additionally, the proposed method outperformed a recent study that employed the same dataset with type-1 and type-2 fuzzy systems, demonstrating improvements in all the parameters and the same for the F-score. This proves the effectiveness and competitiveness of the proposed approach under complex and uncertain conditions in the X platform.