Polynomial Fitting Acceleration Method Based on Time-Series Segmentation

Polynomial fitting is one of the most important approaches of time-series data compression, and the calculation process involves multiple matrix multiplication and inversion, which lead to high computational complexity. Due to the large number of time series instances, the incoming data need to be high performance and real-time processing. Considering the characteristics of time-series data, especially for the periodic sampled time series data, a polynomial fitting acceleration method was proposed. Reusing the intermediate calculation results, both of polynomial fitting methods with fixed and variable segment length were taken to speed up the polynomial fitting of time-series data. Experimental results show that the proposed method can achieve a speedup of 28x and 17x for fixed length segmentation and variable segmentation respectively.