應用機器學習演算法於乳癌資料分析之研究

Application of Machine Learning Methods in Analysis of Breast Cancer Data

近年來，隨著人工智慧與資訊科技的快速發展，應用大數據方法於醫療資料的分析乃是目前醫療產業發展的重點。目前乳癌乃是我國婦女容易罹患的疾病之一，且近年來乳癌的死亡率也逐年攀升。由於有許多因素可能會影響罹患乳癌的判斷，而造成醫療人員在診斷過程的困擾而無法及時進行判斷。為此，本研究藉由特徵選取方法找出重要特徵屬性，並利用決策樹、倒傳遞類神經網路及支援向量機三種機器學習演算法針對乳癌資料進行預測效能的比較分析。最後，本研究採用UCI威斯康辛乳腺癌資料集，進行實證分析。研究結果顯示，當只採用4個屬性時，決策樹及支援向量機的準確率均高達96.19%，因此這些分析模型皆具有良好的預測結果。同時，本研究的特徵選取方法能有效降低資料特徵屬性的個數，同時維持高準確度的預測結果，因而對於醫療人員在進行罹患乳癌的診斷過程具有實務層面的參考價值。

Due to the rapid development of artificial intelligence and information technology in recent years, the analysis of medical data by applying the big data methods is the main development direction of medical industry. Breast cancer is one of the diseases that women are prone to, and the mortality rate of breast cancer has been increasing in recent years. However, many factors will influence the breast cancer judgement in the diagnosis process. Therefore, this study uses the feature selection method to find out the important attributes and uses three machine learning algorithms such as Decision Tree, Back Propagation Neural Network and Support Vector Machine to compare the predictive effectiveness of breast cancer data. This study uses the UCI Wisconsin Breast Cancer Data Set for empirical analysis. The results show that when there are only 4 attributes been used, the accuracy of both the Decision Tree and the Support Vector Machine are 96.19%. Therefore, these analytical models have good prediction results. In addition, the feature selection method in this study can effectively reduce the number of attributes and maintain the prediction results with high accuracy.