In excretion care in nursing homes, it is difficult to change diapers at the appropriate time for each patient due to a lack of manpower, and checking are carried out by regular patrol. Improving the efficiency of this confirmation is one of the issues, for that, a system that equips temperature, humidity and gas sensors into the diaper to detect excretion has been proposed. However, false positives due to various factors have become a difficult problem. In order to identify these variation factors, in this research we focused on the opening of the diaper to confirm the excretion and analyzed the signal patterns. The results showed a clear distinction between before and after opening the diaper, suggested the possibility of using machine learning to discriminate between diaper opening and defecation.