This paper presents a hybrid algorithm between Differential Evolution (DE) and K-means for continuous optimization. This algorithm includes the same operators of the original version of DE but works over groups previously created by the k-means algorithm, which helps to obtain more diversity in the population and skip local optimum values. Results over a large set of test functions were compared with results of the original version of Differential Evolution (DE/rand/1/bin strategy) and the Particle Swarm Optimization algorithm. The results shows that the average performance of the proposed algorithm is better than the other algorithms in terms of the minimum fitness function value reached and the average number of fitness function evaluations required to reach the optimal value. These results are supported by Friedman and Wilcoxon signed test, with a 95% significance.