A novel multidimensional fitness function discrete particle swarm optimization algorithm is proposed to optimize analog test point selection. The proposed method uses fault isolation rate and the number of test points to formulate a multidimensional fitness function to search the global minimal test point set, and an elitist set is used to get more than one possible best solution in the described approach. The efficiency of the proposed method is proven by the same experiments used to verify other methods for optimal test points. Results show that the proposed algorithm in this paper cannot only reduce the computation complexity but also shorten the time consumption. It is particularly useful for large-scale analog circuits.