5. Conclusion and future work
In this paper, we proposed NNCS, an improved CS algorithm with the nearest neighbour strategy and the probabilistic mutation strategy. The nearest neighbour strategy helped the proposed algorithm to generate new solutions learning from their nearest neighbour solutions instead of the best solution obtained so far. In the nearest neighbour strategy, we used a solution-based and a fitness-based similar metrics to select the nearest neighbour solutions. The probabilistic mutation strategy was used to control the solutions learn from the nearest neighbour solutions in partial dimensions instead of all of them in conventional CS. Moreover, our experiments indicated the improvement in effectiveness and efficiency of the nearest neighbour strategy and the probability mutation strategy. The results also revealed that the advantage of NNCS over CS was overall steady as the dimension of problem increases. Additionally, compared with NNCS-S, NNCS-F is a recommended algorithm in terms of solution accuracy, convergence speed, and the convenience of similar metric. The proposed algorithm has the following unique characteristics:(i)in LFRW, it utilizes the non-wheeltopology instead of wheel topology where the best solution influences all other solutions; (ii) it makes the best solution participate in searching according to Eq. (14); (iii) it uses a solution-based and a fitness-based similar metrics to select the nearest neighbour solutions, respectively; (iv) it overall brings solutions with higher accuracy and faster convergence speed; (v) it can deal with high-dimensional optimization problems in terms of scalability study.
