Evaluation of Data Mining Categorization Algorithms on Aspirates Nucleus Features for Breast Cancer Prediction and Detection

With the development of technology the use of Computer Aided Diagnosis has become a key for breast cancer diagnosis. It is important to increase the accuracy and effective of such systems. The concept of data mining can be applied on the data gathered through such systems for prediction and prevention of breast cancer. In this research, we have conducted the comparison between seven classification algorithms with the help of WEKA (The Waikato Environment for Knowledge Analysis) tool on the 569 instances (10 nucleus attributes) of data with two classes Malignant(M) and Benign (B) of breast cancer aspirate cells. Furthermore the influence of each attribute on prediction was evaluated. The accuracy of these algorithms was above 91% with the highest value of 94.02% for random forest and the predictive power of conclave points was highest whereas lowest was of Fractal Dimension.