A lazy associative classifier for time series

摘要: Association rule mining that mainly focuses on symbolic items presented in transactions has attracted considerable interest since a provides concise and intuitive description of knowledge. However, time series is sequence data typically recorded temporal order at fixed intervals time. In to rules the context data, aggregate approximation (SAX) representation could discretize real-valued high-dimensional into segments convert each segment symbol applied this paper. On basis, modified CBA algorithm proposed discover Class Sequential Rules (CSRs) make final prediction first. Then we propose new lazy associative classification method, which computation performed demand driven basis. This contrast rule-based methods like generate excessive number rules, but still unable cover some test with discovered rules. Various experimental results show our for can be interpretable competitive current state-of-the-art algorithm. addition, four different select mined CSR(s) are carrying out classification.