Cancer is one of the diseases that has so far claimed many lives. Recorded in 5 years from 2012 data the International Agency for Research of Cancer (IARC) released more than 14 million people with cancer and 8.2 million of them died of cancer suffered. From these data the most common type of cancer is breast cancer, which is 19.2% of all 14 million more cases. Records related to patients and many types of cancer are carried out in the medical world. The data is increasing and will only become garbage if it cannot be used as new knowledge. Data mining is a field of science that answers the challenges of many data. Classification is part of data mining that allows the creation of new information and knowledge from past data. One of the best and proven classification techniques used is naive bayes. From the 2016 study, naive bayes had the best performance for the classification of breast cancer. Large datasets with many attributes do not guarantee the performance of the algorithm will be better. One process of improving algorithm performance is by selecting features. Gain ratio is the development of an information gain algorithm that is proven to be reliable and can handle high-dimensional data. This study proves that the use of gain ratio feature selection algorithm can improve the performance of Naive Bayes in the classification of Wisconsin Cancer Cancer dataset. Naive Bayes performance without feature selection was 92.7% while after feature selection using the accuracy gain ratio rose 4.01% to 96.71%. Keywords: Data Mining, gain ratio, breast cancer wisconsin, naive bayes