2019/11/01 by Shubham Chandak, Kedar Tatwawadi, Chandak, Shubham +9
Computer Science · #Advanced Data Storage Technologies #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Parallel Computing and Optimization Techniques #Signal Processing (eess.SP) #Time Series Analysis and Forecasting #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1911.00208
openalex publication_date 2019/11/01 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Time series data compression is emerging as an important problem with the\ngrowth in IoT devices and sensors. Due to the presence of noise in these\ndatasets, lossy compression can often provide significant compression gains\nwithout impacting the performance of downstream applications. In this work, we\npropose an error-bounded lossy compressor, LFZip, for multivariate\nfloating-point time series data that provides guaranteed reconstruction up to\nuser-specified maximum absolute error. The compressor is based on the\nprediction-quantization-entropy coder framework and benefits from improved\nprediction using linear models and neural networks. We evaluate the compressor\non several time series datasets where it outperforms the existing\nstate-of-the-art error-bounded lossy compressors. The code and data are\navailable at https://github.com/shubhamchandak94/LFZip\n