Identification of potential classes in procedural code using a genetic algorithm

We present a novel approach for discovering and suggesting classes/objects in legacy/procedural code, based on a genetic algorithm. Initially, a (procedures-accessing-variables) matrix is extracted from the code and converted into a square matrix. This matrix highlights the variable-relationships between procedures and is used as input to a genetic algorithm. The output of the genetic algorithm is then visually encoded using a heat-map. The developers can then (1) either manually identify objects in the presented heat-map or (2) use an automated detection algorithm that suggests objects. We compare our results with previous work.