Frequent pattern mining is a core task in the field data mining, typically relying on the assumption that patterns must be precisely identical. However, in many real-world domains such as social sciences, healthcare, and geology, similarity between objects is often approximate rather than absolute. Similarity measures offer a flexible alternative to strict equality, enabling the discovery of more meaningful patterns. Yet, mining similar frequent patterns often leads to two significant challenges: an overwhelming number of results and high computational cost. Mining frequent closed similar patterns addresses both issues simultaneously while preserving the essential information in the data. In this paper, we systematically evaluate three popular similarity measures—Jaccard, Dice Coefficient, and Kulczynski—in the context of similar pattern mining, by enhancing the closed pattern mining algorithm DCI_Closed. Experiments were conducted on four different databases. Two set of experiments were performed: the first fixes the similarity threshold while varying the minimum support threshold, and the second fixes the minimum support threshold while varying the similarity threshold. Results show that the Jaccard measure offers the fastest performance, owing to its low computational cost and simple formula. In contrast, the Kulczynski measure discovers the most significant number of patterns but consumes the most memory and time. The Dice Coefficient produces balanced results in terms of both the number of patterns and execution time. This study offers insights into the impact of different similarity measures on mining frequent closed similar patterns.