The Application of Data Mining Techniques to the Prognosis of Breast Cancer Patient with Bone Metastasis

• Main Authors: Chen, Howen, 陳鶴文
• Other Authors: Roan, Jinsheng
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/75142859503767574708

碩士 === 國立中正大學 === 資訊管理學系暨研究所 === 99 === According to the statistics from the Department of Health, Executive Yuan of Taiwan in 2000, women died from breast cancer accounted for 3.85% of all cancer mortality. Because of the prolongation of life expectancy, the aging population and the increased number of cancer patients, the skeletal system became one of the common sites metastasized by malignancies of breast, prostate, lung, thyroid gland and kidney. In the previous literature, bone metastases may occur in up to 70% of patients with advanced breast or prostate cancer, and in 40% of other cancers. This study constructed the survival prediction model of breast cancer patients with bone metastasis, by means of decision trees, neural network technology and logistic regression. Those data in this report came from 107 patients at one medical center in Northern Taiwan from 2005 to 2010. In this research, the best model established by data mining for predicting the survival of breast cancer with bone metastases is neural network technology. Its accuracy and ROC curves are 84.55% and 89.5%, respectively. The second-best and the last model are decision trees and logistic regression, whose accuracies are 83.37% and 79.21%, and the ROC curves are 77.1% and 75.3%. These three models can be used clinically to predict the final survival outcomes of breast cancer patients with bone metastases, and meanwhile anticipated to provide physicians with reference and recommendations for evaluation of prognosis and therapeutic efficacy.