A Temporal Data Compression Based on Dual Autoregressive Prediction Method

A temporal data compression method based on rebuilding the predictive model is proposed. This method employs a dual autoregressive prediction model to forecast the temporal data, storing only the residuals between the predicted and actual values, along with the coefficients of the prediction model. During decoding, the prediction model coefficients are used to reconstruct the prediction model to generate predicted values, which are then combined with the residuals to recover the original values. Since the residuals are significantly smaller than the original values, they can be quantized using fewer bits, while the prediction model coefficients, once stored, do not need to be stored repeatedly over time.Experimental results on eight real datasets of varying sizes and types demonstrate the excellent compression performance of this method.