A hybrid approach to automatic programming for the object-oriented programming paradigm
Nelishia Pillay, Caryl K. A. Chalmers · 2007
This paper describes a first attempt at automatically inducing object-oriented programs. The system implemented for this purpose uses a rule-based expert system to firstly derive an object-oriented design (OOD) for the problem instance. Input to the system is a description of the entities in the given domain and their corresponding behaviors. The induced OOD is then input to a genetic programming (GP) system to automatically generate the programs representing the classes. Each class program is evolved in an internal representation language. In order to test the system a translator was developed to convert the evolved class programs to Java. The hybrid system was used to induce object-oriented programs typical of a first course on object-oriented programming (OOP). The paper presents a comparison of the evolved programs to that written by programmers.