Current Proceedings on Technology

Yazarlar: Mahmut Kaya, Oktay Yıldız, Hasan Şakir Bilge

Konular:-

Anahtar Kelimeler:Breast cancer,Missing data imputation,KNN,Naïve Bayes,Mammographic Mass

Özet: Cancer is one of the most mortal diseases in the world. Breast cancer is the second leading cause of death in women. Mammography is a method which is used to detect breast cancer in the initial stage. It helps physicians about in their decisions whether biopsy is necessary or not with respect to tissue shape, border and density. According to researches, 70% of biopsies have done without a need. Because of its cost and complications, it is essential to decide whether biopsy is necessary or not. To achieve this aim, lots of machine learning algorithms are developed to help medical diagnosis in literature, but data sets include some missing values in many real world tasks. These missing values adversely affect classifier performance. Our approach is to impute missing values with k Nearest Neighbor algorithm (kNN) and Naïve Bayes. Then, the performance of the system is evaluated by kNN and Naïve Bayes classifiers to detect breast cancer. Our proposal is measured by performance criteria such as accuracy, sensitivity, specificity and ROC analysis. With this approach, 95 out of 131 missing data which is 9.89% of all data are filled. The experimental results on Mammographic Mass database demonstrate the effectiveness of our proposal with 82.49% accuracy while 81.69% accuracy is obtained without any imputation using same training and test sets. It can be concluded that this approach can also be used for other medical diagnosis problems with high accuracy.

---