AN IMPROVED MULTI-VERSE OPTIMIZER ALGORITHM FOR MULTI-SOURCE ALLOCATION PROBLEM

Multi-verse optimizer (MVO) algorithm is a nature-inspired algorithm for solving single-objective optimization problems. MVO algorithm has many advantages, including few parameters, excellent performance, fast convergence, and low resource consumption. However, it is easy to fall into the local optimum condition and its fineness is not enough during processing multi-resource allocation in Web-based media presentation. Herein, this work proposes an improved MVO (abbreviated as RISEMVO) algorithm. The wormhole existence probability and the travelling distance rate were modified to improve the exploitation capability, and the strategy of revolving around the best universe was added to improve the exploitation and exploration capabilities. Moreover, the jumping o f local optimal strategy was added. In order to reflect the performance differences more appropriately between RISEMVO and MVO algorithms, we firstly tested them with test functions used by the original authors of MVO algorithm. RISEMVO algorithm performs best in 29 test functions compared with standard MVO algorithm and other four commonly used algorithms. We applied RISEMVO algorithm to multi-resource allocation in Web-based media presentation, which enables the maximum utilization of the system and outperforms other 5 algorithms. These results demonstrate the advantages of RISEMVO algorithm in most test functions and in solving the multi-resource allocation problem.