Thesis (M.Sc.) - Cairo University - Faculty of Engineering - Department of Systems and Biomedical Engineering

Data mining effective applications supply good health care knowledge that could be utilized in supporting of clinical decision making. Breast cancer is the most common form of cancer among women and is the second leading reason of cancer death. A risk factor is anything that influences the possibility of obtaining a disease as cancer. The goal of this study is creating an economic method for the early detection of breast cancer with no pain to patient, by analysis of risk factors. Two approaches were followed where in the first approach direct detection has been done between the three cases, benign, malignant and normal by applying decision tree and random forest. In the second approach, indirect detection has been done; it comprises two phases. In the first phase, the detection would be between two categories of normal and tumor cases, applying decision tree and random forest. The second phase was the detection between benign and malignant cases; decision Tree, random forest, K-means clustering and apriori were applied. It is to be noted that in classification and clustering techniques, risk factors were ranked by two different feature selection methods: Fisher linear discriminant and minimal redundancy maximal relevance criterion. In classification techniques 10 folds cross validation method was utilized to decrease the bias of results

Issued also as CD