SUMMARY
Data tables have been widely used for storage of a collection of related records in a structured format in many mobile applications. The lossless compression of data tables not only brings benefits for storage, but also reduces network transmission latencies and energy costs in batteries. In this paper, we propose a novel lossless compression approach by combining co-clustering and information coding theory. It reorders table columns and rows simultaneously for shaping homogeneous blocks and further optimizes alignment within a block to expose redundancy, such that standard lossless encoders can significantly improve compression ratios. We tested the approach on a synthetic dataset and ten UCI real-life datasets by using a standard compressor 7Z. The extensive experimental results suggest that compared with the direct table compression without co-clustering and within-block alignment, our approach can boost compression rates at least 21% and up to 68%. The results also show that the compression time cost of the co-clustering approach is linearly proportional to a data table size. In addition, since the inverse transform of co-clustering is just exchange of rows and columns according to recorded indexes, the decompression procedure runs very fast and the decompression time cost is similar to the counterpart without using co-clustering. Thereby, our approach is suitable for lossless compression of data tables in mobile devices with constrained resources.