Support vector machine is chosen as one of the state-of-the-art discriminative methods with a good performance in many applications. Among linear, quadratic, and cubic kernels, we selected linear kernel due to computation cost. We chose the simple k-nearest neighbor algorithm
from instance-based learning algorithms, and decision tree and decision table from rule-based learning algorithms. In addition, Bayes net is chosen as one of generative models
to show its performance in our experiments. We also used Naïve Bayes and multilayer perceptron as classifiers. All the experiments based on these classifiers were carried out using
Weka developed by Machine Learning Group at University of Waikato [22]. We also used Eclipse to build and run a Java program based on the Java classes supported by Weka.