ANALYZING THE BREAST CANCER DATA IN BMC USING THE TECHNIQUES OF ASSOCIATION RULES AND LINEAR DISCRIMINANT

Abstract

Decision making in considered to be one of the main goals of Applied Statistical Analysis or Data Mining, especially in the field of medicine when decisions about criticalmedical procedures concerning patients are to be made regarding clinical diagnosis, surgical operations and treatments. In this study, a medical data related to breast cancer patients, are being analyzed using two statistical techniques. The data represent patients records in Benghazi Medical Center (BMC) admitted in the period (2003 – 2005). The data matrix consists of 16 variables corresponding to a total of 263 patients (cases).

As a primary step of the analysis, the Exploratory Data Analysis (EDA) powerful tools are being used to explore the distribution and behavior of the data variables to provide a better understanding of patient status. The technique of Association Rules is used in this study as an alternative approach, (to the classical Correlation Analysis), to analyze the multiple causal relationships between data variables, due to their specific nature.The second technique is the Fisher Linear Discriminant Analysis (LDA) which is applied as a supervised method to classify data responses with 2-class and k-class nature based on data predictors. As a general goal of this study, it is attempted to provide a decision making statistical support to help medics in the phase of prognosis of new breast cancer patient cases, based on the previous knowledge (disease behavior).