Diagnosis of software erosion through fuzzy logic

Companies have a vast number of existing software systems, which are not immune to software erosion and ageing as a consequence of uncontrolled maintenance over time. Currently, there are several metrics to measure and quantify software erosion, which also recommends some maintenance actions to deal with software erosion. Unfortunately, there are many symptoms at the same time and several possible maintenance actions that could be carried out. As a consequence, this uncertain environment implies that the best set of actions is unknown and cannot be certainly linked to specific detected erosion symptoms. This paper provides a fuzzy rule-based system to address that challenge. The system is divided into two levels: the first one recognizes precise software erosion metrics and provides fuzzy software erosion symptoms; and the second one takes the fuzzy symptoms and finally obtains fuzzy maintenance actions. This system is therefore a decision-making mechanism to select the best set of actions depending on the specific software erosion symptoms. This system has been implemented using the Matlab Fuzzy Logic Toolbox and it was simulated using Simulink.