Breast cancer is a common cancer type among women. With its increasing incidence early diagnosis has become more important. There are a variety of age-dependent methods for early diagnosis of breast cancer but mammography is the most used method. However, the radiologists show considerable variability in how they interpret a mammogram. Therefore, there is need computer-aided decision-making mechanisms for more reliable results. In this scope various machine learning techniques such as support vector machines, multi layer perceptron and decision trees have been used to early diagnosis in recent years. In this study, centroid-based classifiers are examined for the early diagnosis of breast cancer. The most important reason for this preference is centroid classifiers have low complexity and high performance. Experiments were evaluated on Wisconsin, Diagnostic and Prognostic Dataset. Comparisons between centroid classifiers and the orher classifiers have been done and the results have been presented in terms of accuracy and speed. The highest classification accuracy obtained in the experiments is 99.04%. This classfication rate belongs to the centroid based classifier using the Euclidian measurement. Also, centroid classifiers outperform the other classifiers in terms of classification speed.