Techniques to Compress Time-Series Data

This paper juxtaposes two of the techniques to compress time-series data. As the size of time series data being accumulated is likely to soar, data compression has become crucial in a wide range of applications. This has led to a myriad of data compression techniques for time-series data. In our paper, we have taken two of the well known techniques namely Chebyshev Compression, a lossy technique and Gorilla Compression, a lossless technique which to the best of our knowledge, have never been compared and examined under the same setting. Rendering a “right” choice of compression technique for a particular application is very difficult. To address this problem, we present a benchmark evaluation that offers a comprehensive comparison of both the techniques. Gorilla performs better whenever we have consecutive data points having almost same values whereas, Chebyshev gives good results even when there is no correlation in the dataset. In Chebyshev technique, after transformation, the data values which falls below a pre-determined value (known as threshold), are discarded. Hence, an increase in the threshold value increases the compression gain but this comes at the cost of information loss.