How can we efficiently store and dynamic tensor data? A large portion of real-world data composes of stream data that are in a form of time-growing tensors.For example, air quality measurements are tensor stream data that consists of several sensory data at several locations over time. Such data, accumulated over time, consumes a lot of resources. We need a way to efficiently store tensor data generated in real-time that allow us to effectively retrieve the data for monitoring abnormalities. We propose a dynamic Tucker decomposition method for efficient storage and analysis of multi-dimensional streamline data. Assuming that tensors come in unit blocks, we slice the blocks into matrices and decompose them via singular value decomposition (SVD). The SVDs of sliced matrices, instead of the raw tensor, are stored to reduce space. When data analysis is needed at a certain time interval, involved SVDs of sliced matrices are retrieved to undergo Tuckerdecomposition. The factor matrices and core tensor of the decomposed results can be then used for further data analysis. We compare our proposed method with the existing method and show that our method is more low space cost and analytical. We also apply our method to detect ad-normal air conditions using real-world airquality data.
