Correction of Logical Errors in C programs using Genetic Algorithm Techniques

In this paper, we propose a logical error correction system for C programs. Although there has been considerable research in the field of error detection, there have been few advances in logical error correction. Our solution combines statistical control flow techniques for error detection and genetic programming techniques for error correction. We assume that logical errors are confined to expressions in the input program. In this context, the expressions are mined and statistically ranked on the basis of their error relevance. This initial data forms the input to the genetic algorithm that generates new and improved expression sets for the input program. A main genetic algorithm controls the evolution of expression sets while the individual expressions are modulated by genetic programming techniques. These sets of expressions are tested with the standard test bench and their fitness is declared to the genetic algorithm. The genetic algorithm produces new generations of expression sets until at least one set passes the testing criteria. This algorithm can also be extended to debug codes in other programming languages.