An ant colony optimization based feature selection for web page classification

Area of Science:

• Computer Science
• Information Retrieval
• Machine Learning

Background:

• The exponential growth of web content necessitates efficient automated web page classification systems.
• Web page classification relies on numerous features like HTML tags, URLs, and text content, posing computational challenges.
• Improving search engine performance requires optimizing the accuracy and runtime of web page classification.

Purpose of the Study:

• To reduce the number of features used in web page classification.
• To enhance the runtime and accuracy of automated web page classification systems.
• To evaluate the effectiveness of ant colony optimization (ACO) for feature selection in this domain.

Main Methods:

• Employed an ant colony optimization (ACO) algorithm for optimal feature selection.
• Applied established classifiers including C4.5, naive Bayes, and k-nearest neighbor.
• Utilized the WebKB and Conference datasets for experimental validation.

Main Results:

• Ant colony optimization (ACO) significantly improved both classification accuracy and runtime performance.
• The ACO-based feature selection method outperformed traditional methods like information gain and chi-square.
• Demonstrated superior feature selection capabilities of ACO for web page classification.

Conclusions:

• Ant colony optimization (ACO) is an effective technique for feature selection in automated web page classification.
• Feature selection using ACO leads to more accurate and faster web page classification systems.
• The proposed ACO-based approach offers a significant advancement over existing feature selection methods for web data.