Date of Award


Degree Type


Degree Name



Computer Science

First Advisor

Kobti, Ziad


Evolutionary Algorithm, Game Theory Model with MPCA, Knowledge Migration, Migration in MPCA, Multi population Cultural Algorithm, Optimization




Evolutionary Algorithms (EAs) are meta-heuristic algorithms used for optimization of complex problems. Cultural Algorithm (CA) is one of the EA which incorporates knowledge for optimization. CA with multiple population spaces each incorporating culture and genetic evolution to obtain better solutions are known as Multi-Population Cultural Algorithm (MPCA). MPCA allows to introduce a diversity of knowledge in a dynamic and heterogeneous environment. In an MPCA each population represents a solution space. An individual belonging to a given population could migrate from one population to another for the purpose of introducing new knowledge that influences other individuals in the population. In this thesis, we provide different migration strategies which are inspired from game theory model to improve the quality of solutions. Migration among the different population in MPCA can address the problem of knowledge sharing among population spaces. We have introduced five different migration strategies which are related to the field of economics. The principal idea behind incorporating these strategies is to improve the rate of convergence, increase diversity, better exploration of the search space, to avoid premature convergence and to escape from local optima. Strategies are particularly taken from the economics background as it allows the individual and the population to use their knowledge and make a decision whether to cooperate or to defect with other individuals and populations. We have tested the proposed algorithms against CEC 2015 expensive benchmark problems. These problems are a set of 15 functions which includes varied function categories. Results depict that it leads a to better solution when proposed algorithms used for problems with complex nature and higher dimensions. For 10 dimensional problems the proposed strategies have 7 out 15 better results and for 30 dimensional problems we have 12 out of 15 better results when compared to the existing algorithms.