DESIGN AND APPLICATIONS OF SPIDER MONKEY OPTIMIZATION

Abstract

The contribution of this thesis is the proposal of four new SMO algorithms for solving continuous unconstrained and constrained optimization problems with a view to apply them in solving benchmark problems as well as real world optimization problems. The applicability of SMO over non-linear continuous constrained optimization problems is investigated. A new version of SMO called Constrained Spider Monkey Optimization (CSMO) algorithm has been designed by using Deb’s technique for handling constraints. The performance of proposed CSMO has been investigated over the constrained benchmark problems of IEEE CEC sessions 2006 and 2010. In order to assess the competitiveness of CSMO in solving constrained benchmark problems, it has been compared with three state-of-the-art algorithms namely ABC, DE and PSO on various performance metrics. The results have in presented numerically and graphically. The results have also been validated statistically by using a statistical test. In order to further improve the performance of basic SMO, a new Tournament selection based SMO (TS-SMO) has been designed for solving non-linear continuous unconstrained optimization problems. The performance of proposed TS-SMO has been tested over a benchmark set of 46 benchmark problems and results are compared with basic SMO. For comparing the results, various performance metrics have been taken into account to justify the favourable effect of proposed modification. The results have been compared numerically, graphically and statistically. One more modification of basic SMO named as Quadratic approximation based SMO (QASMO) has been designed for solving non-linear continuous unconstrained optimization problems. The performance of proposed QASMO has been tested over a benchmark set of 46 benchmark problems and results are compared with original SMO. For comparing the results, various performance metrics have been taken into account to justify the favourable effect of proposed modification. The results have been compared numerically, graphically and statistically. Also, a new quadratic approximation based CSMO (QACSMO) has been designed. The performance of proposed QACSMO has been investigated over the constrained benchmark problems of IEEE CEC sessions 2006 and 2010 and the results have been compared with ii CSMO on various performance metrics. The results have in presented numerically and graphically. The results have also been validated statistically by using a statistical test. The main objective behind the development of these algorithms is to apply them over real life optimization problems; hence the applicability of proposed algorithms has been investigated over two real life optimization problems of Lennard-Jones problem and Portfolio Optimization problem. Finally, the thesis is concluded with the overall conclusions, limitations and scope of the proposed algorithms. Also, the future scope and new directions for research in this area have been suggested.