Objective This research investigates Breast Cancer real data for Iraqi women, these data are acquired manually from several Iraqi Hospitals of early detection for Breast Cancer. Data mining techniques are used to discover the hidden knowledge, unexpected patterns, and new rules from the dataset, which implies a large number of attributes.

Methods Data mining techniques manipulate the redundant or simply irrelevant attributes to discover interesting patterns. However, the dataset is processed via The Waikato Environment for Knowledge Analysis platform. The OneR technique is used as a machine-learning classifier to evaluate the attribute worthy according to the class value. Results The evaluation is performed using a training data rather than cross validation. The decision tree algorithm J48 is applied to detect and generate the pattern of attributes, which have the real effect on the class value. Furthermore, the experiments are performed with three machine-learning algorithms J48 decision tree, simple logistic, and multilayer perceptron using tenfold cross-validation as a test option, and the percentage of correctly classified instances as a measure to determine the best one from them. As well as, this investigation used the iteration control to check the accuracy gained from the three mentioned above algorithms. Hence, it explores whether the error ratio is decreasing after several iterations of algorithm execution or not. Conclusion It is noticed that the error ratio of classified instances are decreasing after 5–10 iterations, exactly in the case of multilayer perceptron algorithm rather than simple logistic, and decision tree algorithms. This study realized that the TPS_pre is the most common effective attribute among three main classes of examined dataset. This attribute highly indicates the BC inflammation.