RFlow: uma arquitetura para execução e coleta de proveniência de workflows estatísticos

Abstract

Agricultural data related to the reduction of costs of production and improvements in product quality, prediction and control of pests and epidemics and high precision agriculture are produced on a large scale, and in a heterogeneous manner. The data are captured via sensors, UAVs, the web, satellites, mobile devices, among others. This increasing volume of scientific data and the need to manage and share them between geographically dispersed teams has created a demand for new techniques and computational tools. This study presents the RFlow architecture, a set of integrated tools that manages, shares and collects provenance and reproduces scientific experiments based on R scripts and helps to validate the statistical results. The SisGExp application, one of the architectural components, allows not only access to the data and the processes that are transformed in real time but also collects and records prospective and retrospective provenance descriptors concerning the experiment. In addition, the alignment of the research data to statistical results expands experimental reproducibility, providing greater reliability of scientific results