HBC: Combining Lossy and Lossless Hybrid Bilayer Compression Framework on Time-Series Data

Abstract

The popularization and application of the Internet of Things (IoT) technology has brought massive time series data, which puts forward higher requirements for data compression technology. At present, most existing compression methods use only a single lossy or lossless compression algorithm to perform data compression. Furthermore, traditional lossy compression methods usually adopt a fixed error threshold. However, in practical applications, users have different accuracy requirements for time series data in different numerical ranges. In this paper, we design a Hybrid Bilayer Compression (HBC) framework, which consists of a data accuracy-aware lossy compression layer and a data feature-aware lossless compression layer. First, the original time series data is lossy compressed on the top layer of HBC, where the error threshold can be adaptively adjusted according to the user’s accuracy requirements. Then, based on the features of lossy compressed data, we use supervised learning to select the optimal lossless compression algorithm from the algorithm pool to further compress the data. Experimental results show that compared with three state-of-the-art compression algorithms, HBC reduces the storage space by 52.1%, 91.66%, and 92.27%, respectively.