Composite pattern matching in time series

For last few years many research have been taken place to recognize various meaningful patterns from time series data. These researches are based on recognizing basic time series patterns. Most of these works used template based, rule based and neural network based techniques to recognize basic patterns. But in time series there exist many composite patterns comprise of simple basic patterns. In this paper we propose two novel approaches of recognizing composite patterns from time series data. In our proposed approach we use combination of template based and rule based approaches and neural network and rule based approaches to recognize these composite patterns.