Uma arquitetura orientada a serviços para produzir séries temporais a partir de fontes de dados heterogêneas

Abstract

Real-world phenomena and processes can be analyzed using time series, which are collections ofobservations over time derived from different data sources – typically heterogeneous data sources.Nonetheless, the production and analysis of multiple time series derived from heterogeneousdata sources are complex tasks owing to the different data formats, types, and levels, as well asthe lack of data integration and harmonization tools. This work designed a service architectureto integrate heterogeneous data sources and produce time series in different spatiotemporalresolutions. A rules-based approach that is capable of coupling different data sources toservice architecture was established to harmonize the time series. A concept proof applyingthe architecture was performed to analyze rain events from two distinct data sources: rainfallsensors provided by the Brazilian National Center for Monitoring and Early Warning of NaturalDisasters, and social rain signals derived from Twitter activity related to the rain phenomenon.The results show that the service architecture is flexible and technology-independent since itallows dealing with different data sources, coding styles, and language programming.