Monitoring the perceived quality of remotely delivered video calls for lightweight no-reference predictors. IMPMSPSO-FNN trains a fuzzy neural network (FNN) with a particle swarm optimizer that combines goodpoint-set initialization, five subpopulations, five randomly selected update strategies and a shared global best. It was originally reported to yield large gains on a private dataset, but its update rules, particle encoding and fitness function were incompletely specified, and its evaluation used test errors during training. In this study, we specify the method completely and evaluate it with a nested, leakage-free protocol on LIVE-VQA (compression and simulated IP and wireless transmission; 45 content-separated splits), KoNViD-1k and LIVE-VQC (20 splits each). Baselines are the same FNN trained by standard PSO, differential evolution and Adam, as well as support vector regression (SVR) and random forest on the same 60-dimensional features. IMPMSPSO-FNN reached a mean Spearman rank-order correlation (SROCC) of 0.719 on KoNViD-1k and 0.682 on LIVE-VQC; these values were not significantly different from those of the other FNN training methods and were below those of SVR (0.767 and 0.720; significant on KoNViD-1k). Ablations and repeated-seed runs showed no consistent accuracy gain from any optimizer component, although multi-strategy selection prevented occasional stalls. Using partial least squares inputs instead of principal components raised the SROCC by 0.161 on KoNViD-1k. On LIVE-VQA, every feature-based predictor generalized poorly to unseen contents (mean SROCC at most 0.277). For FNN-based video quality prediction, input representation and leakage-free validation matter more than optimizer elaboration

Md Nadim Hosain, Md Daud Ibrahim, Md Imran Hossain, "Multi-Population Particle Swarm Optimization for Fuzzy Neural Network Video Quality Prediction: Complete Specification and Leakage-Free Evaluation" International Journal of Scientific Research and Engineering Development, V9(6): Page(155-174) November - December 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.