Abstract
Time series data compression is emerging as an important problem with thegrowth in IoT devices and sensors. Due to the presence of noise in thesedatasets, lossy compression can often provide significant compression gainswithout impacting the performance of downstream applications. In this work, wepropose an error-bounded lossy compressor, LFZip, for multivariatefloating-point time series data that provides guaranteed reconstruction up touser-specified maximum absolute error. The compressor is based on theprediction-quantization-entropy coder framework and benefits from improvedprediction using linear models and neural networks. We evaluate the compressoron several time series datasets where it outperforms the existingstate-of-the-art error-bounded lossy compressors. The code and data areavailable at https://github.com/shubhamchandak94/LFZip