Optimized association rules using objective function for mammography image classification

Image mining is more than just an extension of data mining to image domain. In recent years, the concept of utilizing association rules for classification has emerged. This approach proved often is more efficient and accurate than traditional techniques. This paper presents the concept of association rule mining and applied to the problem of mammogram image classifications. Association rules are obtained using Apriori algorithm. Authors propose graph theory based objective function to optimize association rules such that graph generated by the optimized rules is simple graph with simple walk. The proposed algorithm is tested on mammogram images for classification of images into benign and malignant classes. Through experimentation, it is estimated that, with and without optimization of association rules accuracy is 85% for malignant and 95% for benign class. The average accuracy is 90%. The propose technique reduces the time and space complexity associated with calculating optimized rule while maintaining the classification accuracy.