Content area

Full text

Abstract

Machine Learning algorithms have been widely used to solve various kinds of data classification problems. Classification problem especially for high dimensional datasets have attracted many researchers in order to find efficient approaches to address them. However, the classification problem has become very complicated and computationally expensive, especially when the number of possible different combinations of variables is so high. Support Vector Machine (SVM) has been proven to perform much better when dealing with high dimensional datasets and numerical features. Although SVM works well with default value, the performance of SVM can be improved significantly using parameter optimization. We applied two methods which are Grid Search and Genetic Algorithm (GA) to optimize the SVM parameters. Our experiment showed that SVM parameter optimization using grid search always finds near optimal parameter combination within the given ranges. However, grid search was very slow; therefore it was very reliable only in low dimensional datasets with few parameters. SVM parameter optimization using GA can be used to solve the problem of grid search. GA has proven to be more stable than grid search. Based on average running time on 9 datasets, GA was almost 16 times faster than grid search. Futhermore, the GA's results were slighlty better than the grid search in 8 of 9 datasets.

Keywords: support vector machine, genetic algorithms, parameter optimization

Copyright © 2016 Universitas Ahmad Dahlan. All rights reserved.

1. Support Vector Machine Classifier

Classification is a supervised learning technique which learns a function from training data set that consists of input features/attributes and categorical output [1]. This function will be used to predict a class label of any valid input vector. The main goal of classification is to apply machine learning algorithms to achieve the best prediction accuracy [2].

Classification problem can be viewed as optimization problem where the goal is to find the best model that represents the predictive relationships in the data [3]. Other than the wellknown classical data mining techniques such as naive Bayes, decision tree, rule induction, etc., Support Vector Machine (SVM) has gained more attention and has been adopted in data classification problems in order to find a good solution. [4] reported that SVM has been proven to perform much better when dealing with high dimensional datasets.

SVM is...

Show less