Algoritmos paralelos para transformações de padrões ordinais

Abstract

This work presents the development and validation of parallel algorithms for transforming time series into ordinal patterns, focusing on computational efficiency in high-performance environments. The proposed methodology is based on adapting the Bandt e Pompe (2002) method by incorporating parallelism techniques to address the demands of large datasets, such as those generated by IoT devices. The algorithms were implemented in C++ using the OpenMPI library and tested on a high-performance computing cluster, utilizing a real dataset from the Australian energy market. Experimental results demonstrated significant performance gains, highlighted by increased speedup and reduced execution times. Furthermore, the permutation entropy and statistical complexity metrics were preserved with integrity, underscoring the robustness of the method. This study contributes to the efficient application of parallel processing techniques in time series and provides a relevant benchmark for future research