Abstract:
In supervised learning, classifiers are trained with data consisting class labels to solve real-world classification problems. Decision trees, random forest, naïve Bayes, Bayesian networks, K- Nearest neighbourhood logistic regression, artificial neural networks and support vector machines are some of the most popular classification techniques among the class of popular classifiers available for researchers. Although each one of these techniques has their own strengths in dealing varied real-world problems, they have inherent problems too. No classification technique can be universally applied for all real world applications. In the past, several researchers tried to understand the behaviour of these techniques by applying to different areas of research. In the present paper, we experimented with training and test datasets used by some of the researchers to get better understanding on the behaviour of the above classifiers. Study reveals the inadequacies of some of the techniques and superiority of support vector machines and logistic regression over other tools used.
