Özet:

Breast cancer is one of the most common cancers in women worldwide and early detection can be life-saving. This study aims to develop an accurate and reliable model for breast cancer diagnosis by focusing on the Wisconsin Breast Cancer Diagnosis (WDBC) dataset. In the first stage, feature selection was performed using the Minimum Redundancy Maximum Relevance (MRMR) method. The method is used as an effective tool in the field of data mining and feature selection. With MRMR, the importance of the features is ranked and only the significant ones are used. Feature selection improves performance while reducing the complexity of the model. Then, these features selected by MRMR are classified by a Gated Recurrent Unit (GRU) based neural network model for breast cancer classification. The GRU is capable of handling one-dimensional feature series and gives effective results in complex classification problems. The results show that this innovative approach is highly successful in breast cancer diagnosis. In the evaluations, 98.28% for accuracy metric, 98.59% for precision metric, 98.59% for sensitivity metric, 97.67% for specificity metric and 98.59% for f-score metric were obtained. The results show that the method can help specialists in clinical practice. It is understood from the results that the proposed approach has important advantages such as accessibility for all segments of the society, fast and high accuracy even in simple systems.

Anahtar Kelimeler: