An Empirical Analysis of Genetic Algorithm with Different Mutation and Crossover Operators for Solving Sudoku

Prospective optimization tools such as Evolutionary Algorithms (EAs), are widely used to tackle optimization problems in the real world. Genetic Algorithm (GA), one of the instances of EAs, has potential research avenues of testing its applicability in real-world problems and improving its performance. This paper presents a study on the capability of the Genetic Algorithm (GA) to solve the classical Sudoku problem. The investigation includes various mutations and crossover schemes to unravel the Sudoku problem. A comparative study on the performance of GA with these schemes was conducted involving Sudoku. The findings reveal that GA is ineffective to deal with the Sudoku problem, as compared to other classical algorithms, as it often fails to disengage itself from some local optimum condition. On a positive note, GA was able to solve the Sudoku problems much faster, only the Sudoku had very few unfilled elements. A critical appraisal of the observed behavior of GA is presented in this paper, covering combinations of two mutations and three crossovers schemes.