Correlation-Based Feature Selection for Association Rule Mining in Semantic Annotation of Mammographic Medical Images

• Published in: Information Retrieval Technology pp. 482 - 493
• Main Authors: Abubacker, Nirase Fathima, Azman, Azreen, Doraisamy, Shyamala, Azmi Murad, Masrah Azrifah, Elmanna, Mohamed Eltahir Makki, Saravanan, Rekha
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2014
• Series: Lecture Notes in Computer Science
• Subjects: association rule mining; Correlation-based feature selection; mammographic image classification; multivariate filters; semantic annotation
• ISBN: 9783319128436; 3319128434
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-12844-3_41

Mining of high dimension data for mammogram image classification is highly challenging. Feature reduction using subset selection plays enormous significance in the field of image mining to reduce the complexity of image mining process. This paper aims at investigating an improved image mining technique to enhance the automatic and semi-automatic semantic image annotation of mammography images using multivariate filters, which is the Correlation-based Feature Selection (CFS). This feature selection method is then applied onto two association rules mining methods, the Apriori and a modified genetic association rule mining technique, the GARM, to classify mammography images into their pathological labels. The findings show that the classification accuracy is improved with the use of CFS in both Apriori and GARM mining techniques.