Uma proposta de protocolo de apoio à pesquisa em computação

Abstract

In various research fields, the introduction of a new technique is usually accompanied by a performance analysis in order to prove its effectiveness against existing ones. In this direction, special attention should be paid to the performance study being carried out so that the results obtained can be attested. However, scientific articles published in events and journals, show that researchers spend a considerably larger portion to research time design and conducting his experiments than reporting the results obtained. In consequence, it is observed that, in many cases, performance data are presented with an improper validation or no validation at all. As results, many papers are difficult to understand and to reproduce. In other hand, the adoption of rules, defined by guidelines or protocols, to develop the main topics of the scientific method is strongly recommended in other areas of knowledge. In the Computer Science field such initiatives still are lack and in small number. In this work, the object of study is to propose a set of guidelines to support the Computing Science research, taking into account the principles of transparency and reproducibility. The lack of related work in the field of Computer Science with great acceptance implies that initiatives that protocols widely used in the Health Sciences fields guided the conception of the proposed guidelines. Particularities of experiments aimed to comparing two or more objects of study, a very popular type of research in computing, were also considered. The preliminary version of RODA was presented considering the concepts of Computer Research, Statistics and Guidelines, similarly to the works identified in a systematic review. Thus, RODA as validated according to the research subjects and adequacy criteria by the faculty of the UFPel Computing Science courses, followed by the writing of its 25 orientations of final version.