以環境適應值為基礎之鯨魚演算法應用於肝病資料集分類

Fitness-based Whale Optimization Algorithm Applied to the Classification of Liver Disease Dataset

隨著現代人生活及飲食習慣的改變，罹患肝疾病者比率一直位居國人十大死因的前十位，對於國人的健康造成了莫大的影響。因為罹患肝病者在患病初期常未有特別明顯之特徵，因此一般人無法容易察覺自身是否罹患了肝病，致常輕忽後造成肝病的惡化。故若能在早期診斷出肝病前期的症狀，並給予適當的衛教及治療，將可延緩成為嚴重肝硬化疾病的可能。隨著近年來智慧計算之快速發展，科學上逐漸將智慧計算應用於各式學科中，因此本文以環境適應值為基礎之鯨魚演算法來提升原始鯨魚演算法的計算效率，並採用13種標準案例來作為驗證。進一步地，本文採用了UCI的ILPD肝病資料集，資料集內有583筆資料，11項資料屬性。分析方法中除了原始與改良式鯨魚演算法外，亦採用Weka探勘軟體內建的J48決策樹、純樸貝氏、貝氏決策、多層感知機四種方法，來比較分類法對於ILPD肝病資料集之分類正確率。

As the lifestyle and eating habits of people nowadays changed, liver disease has not only caused great damage to people's health, but also been the top ten among the top ten leading causes of death for a long time. Because liver disease has no particularly significant characteristics at the initial stage, people cannot easily detect whether or not they have liver disease. However, if one can diagnose the pre-liver disease early, and give proper health education and treatment, it can delay it to become a serious liver cirrhosis disease. With the advancement of algorithms on intelligent computing, a variety of algorithms have been gradually applied to various disciplines in science. Therefore, this study proposes a fitness-based Whale Optimization Algorithm (WOA) to enhance the performance of the original WOA. Some experiments are performed using 13 benchmark functions by the proposed WOA and original WOA. Furthermore, this study conducts the data mining analysis of ILPD (Indian Liver Patient Dataset) reported in the UCI machine learning repository. The dataset of ILPD includes 583 instances and 11attributes. To classify the dataset of liver disease, we also use four data mining techniques included in the Weka software: J48, Naïve Bayes, Bayes Net, and Multilayer Perceptron. Moreover, the original and improved WOAs and the four aforementioned algorithms included in the Weka are evaluated for the classification analyses of ILPD dataset to compare the classified accuracy of the liver disease dataset.