| Summary: | 碩士 === 輔仁大學 === 企業管理學系管理學碩士班 === 99 === Prostate cancer has ranked as the seventh major disease men dead in our country for several years. Therefore the prevention and screening of disease is getting more and more important. However, most male patients do positive checks after lesions occurred and miss the opportunity of early detection and treatments. If the development of prostate cancer check can be faster and easier, it will definitely provide better and on time treatments for cancer patients. Due to the fact that prostate specific antigen (PSA) index, can be obtained through blood tests, is highly related to prostate cancer, early and effective defection of prostate cancer may be possible if blood test can serve as the first screening mechanism.
The purpose of this study is to investigate the feasibility of classifying PSA using three commonly adopted data mining techniques, namely, discriminant analysis, logistic regression, and backpropagation neural networks (BPN). In order to demonstrate the effectiveness of the three proposed approaches, classification tasks are performed on 1,888 surveys from a hospital in Taipei. Analytic results demonstrated that BPN has the best classification capability in terms of classification accuracy and misclassification costs and can serve as an important alternative in predicting the PSA index as a diagnostic tool in detecting prostate cancer.
|
|---|
