Recomendação de assets híbridos de software utilizando técnicas de aprendizado de máquina

Abstract

Software reuse is an effective alternative in creating software with quality. It also supports some of the Information Technology (IT) demands when creating new products from existing software, softening the effort and costs when compared to products developed from the scratch. Thanks to the vast availability of resources available to developers and to software houses, software reuse becomes more promising, yet more challenging. In order to make opportunistic reuse a more effective approach, and less dependent on human effort to find the right assets, machine learning techniques have been added to opportunistic reuse processes. This study summarizes a systematic mapping, characterizing how the machine learning area has supported the reuse of hybrid software assets. Therefore, this work reports three experiments: two of them carried out seeking to demonstrate how two techniques identified through a systematic literature mapping can recommend assets represented by the RAS++ language; the third experiment compares the two techniques used based on two metrics also identified through the mapping conducted previously. The results turned out to be inconclusive, and they are discussed later along with the factors that led to such a conclusion.