Relationship among prognostic indices of breast cancer using classification techniques

Abstract The main focus of this article is to identify relationships among prognostic indices for different breast cancer groups, using classification algorithms in the field of data mining. Typically, data mining algorithms are used to discover the hidden structure of data. This study is conducted using data from 624 patients suffering from breast cancer in Iran. The information utilized as features includes age, number of involved lymph nodes, location of tumor, vascular involvement, perineural involvement, status of progesterone receptor, status of estrogen receptor, status of HER2/neu receptor, P53, and different breast cancer group factors. The goal is to build different classification models in order to discover patterns from the used dataset. These patterns show the main rules and relationships among different factors. We mainly employed two different learning algorithms, the decision tree learning and a rule-based algorithm. We prepared several experiments to evaluate the impact of different factors in breast cancer diagnosis. Our experimental results show significant relationships between different prognostic indices in the used breast cancer dataset.