Abstract: The Bees Algorithm (BA) is a new population-based optimization algorithm inspired by the foraging nature of bees. In the basic version of the Bees Algorithm, the algorithm performed a combination of neighborhood search and global search. However, the current BA has the disadvantage of not fully imitate all physical and social aspect of bees’ nature. In this study, enhancements to the BA will be introduced as Memory-based Bees Algorithm (MBA) by adding memory (local and global) to two types of bees to make the algorithm more natural. The results of comparing the proposed Local-MBA, global-MBA and MBA (combination of Local-MBA and global-MBA) are tested using several benchmark functions. They had obtained approximately 59.34, 73.02, 74.9 and 75.44% improvement on mean number of evaluations over the basic BA, respectively. Novel fitness values of two engineering design problems are obtained by applying MBA. The proposed algorithms have great potential to be used in many optimization problems.