癌症自1986年起便居於國人十大死亡原因之首,每年約造成三萬人的死亡,從數據上來看,我們可以發現癌症不但影響病患及家人的生活品質,也造成龐大的工作人年損失和醫療費用。綜觀過去關於癌症國內外之學術研究,大部分研究皆集中於癌症治療方式運用和預後因子存活率的探討等,鮮少針對病人及疾病特性來進行適當治療方式的分析。本研究為能清楚了解癌症病患存活時間與存活影響因素間之關係,因此運用獨立成份分析法結合Cox比例風險模式來建構癌症病患存活時間之預測模式,並實際應用於醫療資料中。由於在模式建構的過程中,當選取解釋變數過多時,常會因為解釋變數間具有高度相關性而造成多元共線性問題,進而導致醫生誤判用藥或治療方式等情形出現,因此本研究提出獨立成份分析法針對解釋變數進行資訊萃取來解決共線性的問題,並以模擬實驗方式來驗證獨立成份分析法解決共線性問題之有效性。根據研究結果,我們發現運用獨立成份分析法結合Cox比例風險模式,以建構癌症病患存活時間之預測模式,所得之預測結果與實際結果相一致。此外,本研究所提出的獨立成份分析法能有效解決多元共線性的問題。
In Taiwan, cancer stands as top one of the most lethal disease. Not only affect the quality of life of the patient and their families, but also cause a huge medical expenses and years of potential life lost. In order to reduce the incidence of cancer effectively, we try to find out the causes of low cancer rate by analyzing the pattern and availability of survival influence factors. The objective of this research is to investigate factors which influence survival time of cancer patients. Because survival time can be impacted by factors which are highly correlated with each other, to appropriate treatment of medical, the problem of multicollinearity must be solved. Therefore, this research proposes a new solution, Cox proportional hazards model combines independent component analysis method, to eliminate the multicollinearity among explanatory variables. To evaluate the performance of the proposed method relative to alternative approaches, we report one experiment study based on the dataset from Monte Carlo simulation experiments. The result shows that the proposed approach can solve the problem of multicollinearity and the significant effect between survival time and factors with cancer patients.