The cell formation (CF) problem is considered the most essential issue in cellular manufacturing systems (CMS). CF deals with the arrangement of similar parts into groups known as part families (PFs) and organizes machines also into groups, called machine cells (MCs). In the literature, numerous methods, models and algorithms have been proposed and developed to handle CF problems. However, very few studies have dealt with the assessment and comparison of these methods, to identify the most effective. This has provided strong motivation for the study presented here. The present paper focuses on two methods that are used infrequently to form MCs and PFs, and applies them in three strategies: the first is based on the use of a hamming distance only, while the second uses only a self-organization map (SOM). However, the third method applies a hybrid approach based on SOM and hamming distance. The outputs of the selected methods were compared, to select the best one. A set of five benchmark datasets and three performance measures was used for comparison and evaluation. These performance measures are: percent of the exceptional elements (PE), grouping efficiency (GE), and machine utilization (MU). The results refer to the outperforms of the hamming distance in terms of PE, GE and MU for most of the selected benchmark problems.