Efficient Pattern Recognition in Time Series Data

In recent years, Time Series analysis has become one of the most important part of the research area. Classification of time series based on pattern is mainly used in medical and robotics domain. Very large amount of data is generated, which can be stored as time series data. As the size of time series is very large, traditional algorithms are not efficient to recognize and classify patterns extracted from it. To recognize and classify patterns from time series efficiently a new method is proposed in this paper, in which clusters are computed of time series based upon similarity of pattern in them. NN classifier used to classify time series, which uses previously computed cluster center. This method reduces the computations which directly affect the performance of the classification algorithm. The proposed system classifies various time series datasets and is outperforming the previously used tradition methods. For large scale time series, proposed method is implemented on Hadoop framework to achieve high scalability.