Optimizing support vector machine parameters using cuckoo search algorithm via cross validation
Akkawat Puntura, Nipon Theera‐Umpon, Sansanee Auephanwiriyakul · 2016
Support vector machine is one of the most popular techniques for solving classification problems. It is known that the choice of parameters directly affects its performance. This problem can be solved using a search algorithm which is suitable optimization technique for the parameter optimization. In this research, we propose a method to determine the optimal parameters for support vector machines using the cuckoo search algorithm via maximization of the average accuracy from k-fold cross validation. Our experimental results show that the cuckoo search algorithm provides very good convergence rate and outcomes. The comparison between its performance and another population based optimization namely the particle swarm optimization is also performed. It shows that the cuckoo search algorithm yields better convergence rate and outcomes than the particle swarm optimization in most datasets. It implies that the mechanism of cuckoo search algorithm is efficient for this parameter optimization problem and is more effective than the particle swarm optimization in this particular problem.