Simulation of pornography web sites (PWS) classification using principal component analysis with neural network

Abstract

The explosive growth of objectionable web content such as pornography, terrorist and violence had been a serious threat for internet users especially children. Recently content analysis based filtering is being introduced to overcome this problem. In term of the promising result to satisfy the result of web content analysis, features extraction techniques play an important role to extract appropriate features from large volume of web information such as text, image, audio, video etc. In this paper we propose a model of pornography web site classification which mainly based on textual contentbased analysis such as indicative keywords detection. This paper will show that implementation of principal component analysis in back-propagate neural network is capable to classify high similarity illicit web content sufficiently. In this study, we introduce three techniques to implement our Pornography Web Site Classification Model (PWSCM) such as PWSCM with principal component analysis (PWSCM-PCA), PWSCM with only CPBF (PWSCM-CPBF) and PWSCM with integration of CPBF and PCA (PWSCM-CPBF-PCA). We compare the performance of these three techniques by conducting several simulation experiments. From the experiment results, we have found that the proposed model with three different techniques capable to perform efficient identification for illicit web content. Hence this paper will discuss the simulation results of the model with three techniques.