Tcm: A High-Precision Lossy Compression Algorithm for Time Series Data With Flexible Queue-Value Dynamic Grouping

Approximating a series of time-stamped data points with a sequence of line segments while guaranteeing a maximum error is a fundamental data compression problem called Piecewise Linear Approximation (PLA). Segmented Linear Approximation (PLA) is a well-established tool for reducing the size of a time series representation by approximating the time series with a sequence of lines while keeping the error introduced by the approximation within a predetermined threshold. These algorithms can help us process large amounts of information, albeit at the cost of some loss of precision. More precisely, these algorithms need to strike a delicate balance between the maximum acceptable loss of precision and the achievable space savings.For the problem of efficient compression of time series data, we propose an innovative algorithm based on dynamic error thresholding and grouping of similar line segments (Tcm). The algorithm dynamically optimizes the error threshold (the maximum error is less than 0.1%) by introducing the simulated annealing technique, and combines the adaptive chunking strategy with the multimodal data preprocessing (sliding average filtering, wavelet denoising, and STL-TSVM decomposition), which greatly improves the compression efficiency of Piecewise Linear Approximation (PLA). Experiments show that Tcm achieves a better balance between compression rate and accuracy: compared with algorithms such as Sim-Piece and Mix-Piece, its compression rate is improved by 28.1% on average, and the compression time is shortened by 30%, which is especially prominent in strongly trending scenarios such as financial time series and industrial sensing. The method provides a highly robust solution for real-time processing of massive time series data.