A Memetic Algorithm (MA) is an evolutionary algorithm that includes one or more local search phases within its evolutionary cycle.Over the last decades,MAs have relied on the use of a variety of different methods as the local improvement procedure.Studies have shown that the local search method employed in a MA has significant influence on its performance.To overcome the restricted theoretical knowledge of a problem and mitigate the effects of an incorrect choice of local search method,we propose an adaptive MA: a memetic algorithm based on k-Nearest Neighbor(k-NN) in this paper.It uses multi-memes(multi-local search methods) and adopts an adaptive method based on k-Nearest Neighbor to choose a local search method at runtime.Finally,experiments on several continuous benchmark problems of diverse complexities show that the new approach is able to provide highly competitive results compared with other algorithm.