“…However, user preference attributes cannot be accurately obtained. MADM-based network selection algorithms, for instance, simple additive weighting (SAW) [3,4], multiplicative exponent weighting (MEW) [5], grey relational analysis [6], order preference by similarity to ideal solution (TOPSIS) [7] as well as analytic hierarchy process (AHP) [8][9][10][11], rely on experiences to get near-optimal selection, but these methods lack strict theoretical proof of optimality. Machine learning-based algorithms in [12,13] utilise reinforcement learning to get network selection, while they depend too much on the accuracy of the data set and the processor performance.…”