סמינר מחלקתי

ON THE CONTRIBUTION OF DESIGN ATTRIBUTES TO THE CLASSIFICATION OF WEBSITES CATEGORIES

Doron Cohen -  M.Sc. student

The Department of Industrial Engineering, Tel-Aviv University, Israel

Abstract:

Web mining is the use of data mining techniques to automatically discover knowledge from the World Wide Web. One of the most challenges applications of web mining is detecting websites by their category and in particular – detecting malicious websites. Developing tools for website's categories detection is often required to be resources expensive as the enormous amount of information found in the web continues to rise rapidly. In this study we present an algorithm which is able to pull websites content and design by a given URL. We have activated the algorithm and processed the information of 450 websites. Then, we examined the websites information in front of their categories taken from "Google's top 1000 sites" and we built a prediction model using J48 tree with 10-folds cross validation. In the first experiment, we show that classification by design and HTML colors features only (with no use of text mining and other sophisticated tools) is able to predict five different websites categories (including malicious). In the second experiment, we show that adding attributes of design to other objective prediction method, can enhance predictability of malicious websites to high accuracy percentage of ~97.8%, with low resources usage and low run time. Furthermore, we have used T-test and we found significance enhancement. We conclude that colors have great importance when predicting websites categories.

This work was performed under the supervision of Prof. Irad E. Ben-Gal and Prof. Shulamith Kreitler

ההרצאה תתקיים ביום שלישי 9/12/14, בשעה 14:00 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב.