Online version of fuzzy rule-based classifiers is applied to detecting the abnormality in the life pattern of elderly people.A fuzzy rule-based classifier consists of a set of fuzzy if-then rules with a real value in its consequent part.The standard type of fuzzy rule-based classifiers are formulated under the assumption that all training patterns are available before a classifier is constructed.In order to accommodate the situation where training patterns are streamed and only a small number of training patterns are available at a time,online fuzzy classifiers are presented where confidence-weighted learning algorithm is used to adapt fuzzy classifiers according to the newly available training patterns.The life patterns of an elderly person are virtually obtained through an occupancy simulator that is based on a set of real observatory data.A series of computational experiments are conducted to examine the performance of the online fuzzy classifiers.