A hybrid Analytical Hierarchical Process and Deep Neural Networks approach for classifying breast cancer
Taysir Hassan A. Soliman, Randa Mohamed, Adel A. Sewissy · 2016
Data mining techniques and multi criteria decision making techniques have been used widely in many areas, such as customer relationship management, medicine, engineering, education, geographic information systems, and recommendation systems. The present study aims to design a hybrid approach based on Deep Neural Networks (DNNs) and multi criteria decision making. DNNs and multi criteria decision making techniques are integrated with Analytical Hierarchy Process (AHP) to improve classification accuracy and deal with large datasets. Three different breast cancer datasets are used for evaluating the performance of the proposed hybrid approach. In most cases, the hybrid approach of applying DNN Backpropagation with three hidden layers and AHP gives better accuracy rate 84.33%, precision 95.9%, recall 86.6%, and F-measure of 90.2% than using one hidden layer or applying DNN Backpropagation with three hidden layers without AHP. In addition, classical classification algorithm are also compared: J48, naïve bayes, and random tree, where they give less results than proposed approach.