A dandelion algorithm (DA) is a recently-proposed intelligent optimization algorithm and shows an excellent performance in solving function optimization problems. However, like other intelligent algorithms, it converges slowly and falls into local optima easily. To overcome these two flaws, a dandelion algorithm with probability-based mutation (DAPM) is proposed in this paper. In DAPM, both Gaussian and Levy mutations can be used interchangeably according to a given probability model. In this paper, three probability models are discussed, namely linear, binomial, and exponential models. The experiments show that DAPM achieves better overall performance on standard test functions than DA.
