Perbandingan Klasifikasi Website Secara Otomatis Menggunakan Metode Multilayer Perceptron dan Naive Bayes

Abstract

World wide web has become a big repository that contains lot of information. As the number of websites continuous growth exponentialy, the need of website classification gains attractions. The human ability to perform manual classifications is increasingly difficult. Website Classification using machine learning technique become more important to do, because the classification process is done automatically.  The classification system begins with the process of collecting information from the home page of the web site(parsing). Home page of a web site is a distinguished page and it acts as an entry point by providing links to the rest of the web site. For each parsing result of the homepage there are process of removing the stop word, stemming and feature selection with tf-idf. The result of this process is a feature that becomes input of machine learning algorithm. In this algorithm there are learning process of input pattern and making of the weight. This weight will be used in the classification process. In this research, the learning process is developed by using multi layer perceptron and naive bayes algorithm. The classification results of each method will be compared. Based on the results obtained, naive bayes have a better accuracy rate than multilayer perceptron. With accuracy of 89% Naive Bayes and MLP has 80% accuracy