A Comparative Study of Multilayer Neural Network and C۴.۵Decision Tree Models for Predicting the Risk of Breast Cancer

Background: Diagnosing breast cancer at an early stage can have a great impacton cancer mortality. One of the fundamental problems in cancer treatment is the lackof a proper method for early detection, which may lead to diagnostic errors. Usingdata analysis techniques can significantly help in early diagnosis of the disease. Thepurpose of this study was to evaluate and compare the efficacy of two data miningtechniques, i.e., multilayer neural network and C۴.۵, in early diagnosis of breastcancer. Methods: A data set from Motamed Cancer Institute's breast cancer researchclinic, Tehran, containing ۲۸۶۰ records related to breast cancer risk factors wereused. Of the records, ۱۱۴۱ (۴۰%) were related to malignant changes and breastcancer and ۱۷۱۹ (۶۰%) to benign tumors. The data set was analyzed usingperceptron neural network and decision tree algorithms, and was split into two atraining data set (۷۰%) and a testing data set (۳۰%) using Rapid Miner ۵.۲. Results: For neural networks, accuracy was ۸۰.۵۲%, precision ۸۸.۹۱%, andsensitivity ۹۰.۸۸%; and for decision tree, accuracy was ۸۰.۹۸%, precision ۸۰.۹۷%,and sensitivity ۸۹.۳۲%. Results indicated that both algorithms have acceptablecapabilities for analyzing breast cancer data. Conclusion: Although both models provided good results, neural networkshowed more reliable diagnosis for positive cases. Data set type and analysismethod affect results. On the other hand, information about more powerful riskfactors of breast cancer, such as genetic mutations, can provide models with highcoverage.

Keywords:

Decision tree , multilayer neural network , breast cancer , data analysis