比較三種資料探勘演算法預測大腸直腸癌病患存活的效能

PREDICTING ACUTE COLORECTAL SURVIVABILITY: A COMPARISON OF THREE DATA MINING METHODS

　本研究案例使用支援向量機(SVM)、決策樹(Decision tree)和邏輯式回歸(Logistic regression)分別建構大腸直腸癌病患之預後模式，最後選擇由決策樹建構的模型為本研究最佳模型(異常的正確率為71.73%，正常的正確率為76.04%)，模型可用於預測大腸直腸癌症患者的最終結果。研究結果將可提供醫院或臨床醫療者先行將病患的最終結果做預出並且做出進行客觀評估決策，做為治療時參考之用，最終目的降低大腸癌的發生率以及死亡率。

　　This study compared SVM and Decision tree and Logistic regression. These three models, including SVM, Decision Trees, and Logistic Regression, were evaluated based on the predictive accuracy rate for test sets. The last select were Decision tree that it had better abnormal accurate rate (71.73%) and normal accurate rate (76.04%) than SVM and Logistic regression. The result of this study can advance to predictive patient outcomes such as patient death, dialysis dependent and recovery renal function for clinical treatment.