DRSM: Data Reduction and Similarity Matching for Time Series Data Streams

Time-series data streams, such as stock market data, sensor network data, and weather data should be analyzed to predict the future trends in the respective applications. The prediction process needs the similarity matching technique which becomes more complex and time consuming as the data set increases. The task of compaction of data and enhancing the performance of similarity matching of time-series data is very important. To achieve this, for a large time-series data streams, first we segment the data stream and condense to small data stream by retaining all the vital features using Multilevel Segment Mean (MSM) technique. After data reduction, similarity matching is performed by comparing the new arriving data objects with the existing data streams. The proposed approach is experimented and found to be efficient and most suitable for similarity matching over the time series data stream