DIFFERENTIAL EVOLUTION OPTIMIZATION METHOD WITH PARAMETER FREE PENALTY FUNCTION AND SPIRAL CLUSTERING

• Main Author: Immanuel, Jason
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/47813
• Institution: Institut Teknologi Bandung

In this final project, the author propose a newly developed metaheuristic optimization method for solving optimization problem with the ability to find not only find a single solution, but all the possible solutions for an optimization problem. This method will use the Differential Evolution Method as the basis of the solver with the assistance of a penalty function for constrained problems and Spiral Clustering to help find all the possible solutions. Differential Evolution is one of the many approximation methods that work in a population basis for solving optimization problems. The solution for a problem is sought by iteratively moving the population towards a value which might as well be the optimum solution. Each stage in the Differential Evolution steps will be explained in this project, as well as how a penalty function such as the Parameter Free Penalty Function can help the population in the Differential Evolution tp converge into a feasible solution and also the clustering done by the Spiral Dynamics to divide a domain into regions that considered might have an optimum solution. All theories discussed in this project are then implemented in the new innovation method which will be called the PFP-SC-DE method using the python language with the pycharm application. After the method had been successfully developed, the PFP-SC-DE method is then used to solve various types of optimization problems, both the optimization that maximize and minimize, discreet and continues problems, also constrained and unconstrained optimization problems, and finally to optimize a portfolio of 5 selected stocks.