Data mining application with case based reasoning classifier for breast cancer decision support

Abstract

Cytology is a complex diagnosis task which requires both expertise and experience of an oncologist for providing the cancer class and stage which is very useful in the therapy and in the surgery intervention. A case based reasoning classifier is developed with specialized agents for recognizing the malignant breast cancer. The proposed application implements a data mining method for the knowledge extraction and discovery by mining a medical database, which contains classified instances characterized by some features extracted automatically from the cytological image of the patient cancer. An original technique is implemented for enriching the retrieving process on the developed CBR system; this technique is based on the combination of global-local similarity measures and fuzzy sets for modeling the unknown response generated from the agents which increase significantly the accuracy of the system. The features selection and weighting is done by a machine learning algorithm. The efficiency of the proposed methodology has been validated through some empirical experiments applied in the cited data set which demonstrates that the developed approach achieves such average accuracies better than the current state-of-the-art approaches.