許永真臺灣大學：資訊工程學研究所顏士傑Yen, Shih-ChiehShih-ChiehYen2007-11-262018-07-052007-11-262018-07-052007http://ntur.lib.ntu.edu.tw//handle/246246/53743在機器學習領域當中，特徵選擇一直以來都是一個重要的課題，尤其以行為辨識(Activity Recognition)而言，我們利用許多不同的感應器來擷取各種大量的資訊，假如能透過特徵選擇的技術來挑選出重要特徵，將有許多好處，例如增快辨識速度、提高辨識準確度等等。本論文提出一個基於正交實驗的特徵選擇法，並以循序收納選擇法(Sequential Forward Selection)為基準，比較並探討此法的優劣與適用性。Feature selection is an important issue in the problem of machine learning. Especially in the domain of activity recognition, many researchers try to make use of multiple heterogeneous sensors and thus receive a large amount of signals. Many features can be extracted, hence feature selection becomes more important. In this thesis, we propose a feature selection method based on orthogonal experimental design and compare this method with equential forward feature selection in terms of the accuracy of the model induced by the selected feature subset versus the number of treatments and the number of selected features.Acknowledgments ii Abstract iii List of Figures ix List of Tables xi Chapter 1 Introduction 1 1.1 Motivation . . .1 1.2 The Goal of Feature Selection . . . 2 1.3 Challenge of Feature Selection . . . 2 1.4 Problem Definition . . . 2 1.5 Thesis Organization . . . 4 Chapter 2 Related Work . . . 5 2.1 Activity Recognition . . . 5 2.1.1 RFID Technology . . . 6 2.1.2 Accelerometer Technology . . . 7 2.1.3 Audio/Video Stream . . . 7 2.1.4 Heterogeneous Sensors . . . 8 2.2 Feature Selection Strategies . . . 8 2.2.1 Filter Approaches . . . 10 2.2.2 Wrapper approaches . . . 11 Chapter 3 Preliminaries . . . 15 3.1 Basic concept of Orthogonal Experimental Design . . . 16 3.2 Orthogonal Arrays . . . 19 3.2.1 The Definition and Some Properties of Orthogonal Arrays . . . 20 3.2.2 Constructions for Orthogonal Arrays . . . 22 3.3 Factor Analysis . . . 29 Chapter 4 Feature Selection . . . 33 4.1 The Wrapper Approach for Feature Selection . . . 34 4.2 Feature Selection Based on Orthogonal Experimental Design . . . 37 4.2.1 Two Properties . . . 37 4.2.2 Strength . . . 40 4.2.3 Complexity Analysis . . . 40 4.3 Feature Selection Based on Iterative Orthogonal Experimental Design . . . 41 4.3.1 The Algorithm . . . 41 4.3.2 Complexity Analysis . . . 44 Chapter 5 Experiment . . . 47 5.1 Data Description . . . 47 5.1.1 3-bit Parity Dataset . . . 47 5.1.2 Audio Data of the ADLs Dataset . . . 48 5.1.3 Emotion Dataset . . . 50 5.1.4 The Madelon Dataset of NIPS Competition . . . 50 5.2 Evaluation and Experimental Results . . . 51 5.2.1 3-bit Parity Dataset . . . 51 5.2.2 Real World Datasets . . . 52 Chapter 6 Conclusion . . . 61 Bibliography . . . 63624887 bytesapplication/pdfen-US機器學習特徵選擇正交實驗Feature SelectionOrthogonal Experimental Designthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53743/1/ntu-96-R94922089-1.pdf