In this paper we propose a local search based improvement heuristic called diversified local search for the traveling salesman problem. We show through computational experiements that this algorithm outperforms tabu search with similar neighborhood structures on large sized traveling salesman problem instances.