This paper presents a method for analyzing time-series laboratoryexamination databases. The key concept of this method is classificationof temporal patterns using multiscale structure matching and a roughset-based clustering method. Multiscale matching enables us to capturesimilarity between two sequences of examinations from both short-termand long-term points of view. The rough-set based clustering techniqueis then applied to classify the sequences according to the relativesimilarity obtained through multiscale matching. In the experiments weshow that this hybrid approach can be used to discover interestingtemporal patterns hidden in the time-series databases
展开▼