Adaptive inertia weight Bat algorithm with Sugeno-Function fuzzy search

Bat algorithm (BA) turns into the most generally utilized meta-heuristic algorithm to solve the different sort of global optimization problems. In the optimization of continuous data, BA experiences one of the prominent difficulties called premature convergence. In order to tackle premature convergence, this study exhibits a new version of BA called Adaptive inertia weight Bat algorithm with Sugeno-Function Fuzzy Search (ASF-BA). The proposed algorithm ASF-BA brings two major adjustments in the standard BA. Firstly, we incorporated an adaptive inertia weight to boost up the velocity rate of bats effectively. Secondly, we replaced the random searching method of standard BA with Sugeno-Function fuzzy search, which used Sugeno-Function decline curves to dynamically adjust the fitness of each bat according to their own experience and experience of their neighbour bats. We compared ASF-BA with several old fashioned and new fashioned optimization algorithms. ASF-BA is also tested against top hybridized and enhanced version of DE algorithms. The CEC 2017 benchmark (30 real parameter numerical optimization problems ), CEC 2017 ( 28 constrained optimization problems) and 19 additional benchmark problems have been used to examine and compare the performance of ASF-BA with other state of the art variants. Contrasted with the existing BA and other leading variants of BA, DE, and PSO on CEC 2017 constrained and numerical benchmarks, the ASF-BA is excellent to the state-of-art variants of BA, DE, and PSO in terms of stability, convergence speed and solution quality. The ASF-BA sets stable support for resolving optimization problems of intelligent and expert systems. Furthermore, we also examined the performance of proposed ASF-BA for the weight optimization of Feed Forward Neural Networks (FFNN) and compared ASF-BA with Back Propagation Algorithm (BPA), BA and PSO. ASF-BA achieved 94 % of maximum accuracy. The experimental outcomes reveal that the suggested algorithm executed especially reliable and effective than the existing state of the art variants.