Learning to Remove Internet Advertisements

This paper discusses improving the methodology introduced in Kushmerick’s paper about learning to remove internet advertisements. The aim is to reduce the model build time as well as the classification time, while increasing the classification accuracy. Our results showed that with careful selection of features, it is possible to significantly reduce the time to build a model (from as much as 20 seconds to 0.09 seconds), as well as the time to classify new instances, while maintaining Kushmerick’s classification accuracy of 97%. Author Keywords Data mining, classification

• Type: Conference
• Year: 1999
• Where: AGENTS
• Authors: Nicholas Kushmerick
• Keywords: AGENTS 1999, Classification Accuracy, Classification Time, Model Build Time, Security Privacy