Afyon Kocatepe Üniversitesi Uluslararası Mühendislik Teknolojileri ve Uygulamalı Bilimler Dergisi

Yazarlar: Erkan ÖZHAN

Konular:Mühendislik

Anahtar Kelimeler:Web information extraction,Machine learning,Classification,Data mining

Özet: The web environment has a format in which information is born, propagated and lived. Information changes morphologically day by day, and with this change, difficulties in reaching the desired meaningful information increase as well as advantages. It is a critical task to reach the desired information in the most efficient way in terms of time, storage, communication and data processing costs. In addition, it can increase the availability of data throughout its life cycle. Classification of the parts of web pages called “layout” can make important contributions to the solution of these problems. In particular, knowing the unnecessary content in these sections can facilitate access to useful and meaningful information and provide a cost-reducing effect. This study focuses on different algorithms in order to improve the process of classifying web page sections with machine learning methods and reveals the improvement results of these algorithms. According to the results, it has been seen that Random Forest and KStar algorithms have process improvement solutions. While the Random Forest algorithm offers 98.4602% correct classification rate, KStar stands out with its speed factor. In the study, especially the performance of tree and entropy-based algorithms were compared and the findings were presented together with the computation times.

---