Developing Clustering Based on Genetic Algorithm for Global Optimization

摘要

Nowadays, databases are widely used over the world. The huge amount of data requires modern methods to make it useful meaning of information, clustering is one of the techniques that collects similar objects then put them in groups. Clustering is an approach appropriate for extracting useful meaning in large database. K-mean clustering is an algorithm characterized by simplicity and easy to implement and provides good results. However, it suffers from being trapped in local optimal solution. Some hybrid between two algorithms aims to combine the advantages of two algorithms to make optimization. In this thesis, we propose applying the same hybrid between k-mean clustering and Differential Evolution (DE) called Clustering based Differential Evolution CDE, but in the proposed method, we use Genetic Algorithm (GA) instead of Differential Evolution to find a globally optimal solution. This proposed method called Clustering based on Genetic Algorithm for Global Optimization (CGAGO), then we compare between them, hi addition, we use a parameter called cluster period to improve k-mean clustering, in terms of rinding the global optimum. Moreover, we test eleven Benchmark functions to validate the proposed method. Experimental results show that the proposed method CGAGO is slightly better and effective than CDE.