結合核心切片反迴歸與蜜蜂演算法於支援向量機之混合式系統，以醫學資料分析為例

A hybrid support vectors machine model with KSIR and HBMO in analyzing medical data

卓盈玠(Ying-Zhieh Cho)

指導教授 ： 白炳豐

國立暨南國際大學/管理學院/資訊管理學系/碩士(2009年)

摘要

群體智慧(Swarm intelligence)是藉由觀察社會型昆蟲(Social insects)的行為模式，找出其有利特性並應用到人類生活中各個領域的學門。目前群體智慧的發展有螞蟻演算法 (Ant Colony Optimization, ACO)、粒子群演算法 (Particle Swarm Optimization, PSO)與基因演算法 (Genetic Algorithm, GA)等。在資料探勘的分類問題中，支援向量機 (Support Vector Machine, SVM)為近幾年廣泛使用且效果不錯的分類方法，但目前支援向量機最常面對兩種問題，一是如何找出輸入的最佳特徵集合，另一為如何找出核心函數C與σ之最適參數。若特徵集合與核心函數的參數選擇不佳，將影響整體的分類結果。為解決上述問題，本文以UCI公開之醫療資料庫為例，提出結合核心切片反迴歸 (kernel sliced inverse regression, KSIR)與蜜蜂演算法 (honey-bee mating optimization, HBMO)於支援向量機之混合式系統，透過因素分析結合核心切片反迴歸，對資料特徵進行縮減與降維動作，以找出輸入的最佳特徵集合。最後利用蜜蜂演算法之群體參數搜尋能力，找出最適參數組並代入支援向量機內進行分類，以建構出一優良的疾病診斷分類模式。
實驗結果發現，透過因素分析結合核心切片反迴歸法，在分類與執行效率上均優於只作正規化之前處理動作。在整體分類效果上，本文所提出蜜蜂演算法結合基因演算法也優於蜜蜂演算法原型的區域搜尋法。因此，本文所提出之模式，為一快速且能正確分類的疾病診斷分類系統。

關鍵字

支援向量機 ； 蜜蜂演算法 ； 因素分析 ； 核心切片反迴歸 ； 疾病診斷

並列摘要

Swarm intelligence is based on observing the collective behavior of social insects and extract characteristics that can be applied to human life domains, such as ant colony optimization (ACO), particle swarm optimization (PSO) and genetic algorithm (GA). This paper proposes a hybrid model which firstly combines factor analysis (FA) with kernel sliced inverse regression (KSIR) for attribute extraction and dimensionality reduction forming the best selected feature subset. Secondly, honey-bee mating optimization (HBMO) is used to solve the problem of parameters settings in support vector machine (SVM) for classification. Results of the medical dataset from the UCI Machine Learning Repository applying the hybrid model show better results than original methods. Thus, the proposed model is an alternative and helpful scheme in analyzing medical data.

並列關鍵字

swarm intelligence ； honey-bee mating optimization algorithm ； support vector machine ； disease diagnosis ； data mining

參考文獻

Abbass, H. A., (2001), “MBO: Marriage in Honey Bees Optimization A Haplometrosis Polygynous Swarming Approach,” The Congress on Evolutionary Computation.

Abbass, H. A., and Teo, J., (2001), “A True Annealing Approach to the Marriage in Honey–Bees Optimization Algorithm,” Memorias del Inaugural workshop on Artificial Life (AL’01).

Andersen, T. L., and Martinez, T. R., (1996), “The Effect of Decision Surface Fitness on Dynamic Multi-layer Perceptron Networks,” Proceedings of WCNN'96 World Congress on Neural Networks., pp. 177-181.

Andersen, T. L., and Martinez, T. R., (1996), “Using Multiple Node Types to Improve the Performance of DMP,” Proceedings of the IASTED International Conference on Artificial Intelligence., Expert Systems and Neural Networks., pp. 249-252.

Bhattacharya, B., Mukherjee, K., and Toussaint, G., (2005), “Geometric Decision Rules for Instance-based Learning Problems,” Lecture Notes Computer Science., pp. 60-69.