|
|
|
 |
|
|
|
|
|
|
Classification of Zen-meditation and Resting EEG Spatial spectral properties by Random Forest=隨機森林應用於禪定與放鬆休息腦電波之頻率空間特性分類 |
|
|
|
| Author |
蔡宗諺 (著)=Tsai, Tsung-yen (au.)
|
|---|
| Date | 2019 |
|---|
| Pages | 241 |
|---|
| Publisher | 國立交通大學 |
|---|
| Publisher Url |
https://www.nycu.edu.tw/
|
|---|
| Location | 新竹市, 臺灣 [Hsinchu shih, Taiwan] |
|---|
| Content type | 博碩士論文=Thesis and Dissertation |
|---|
| Language | 英文=English |
|---|
| Degree | master |
|---|
| Institution | 國立交通大學 |
|---|
| Department | 電控工程研究所 |
|---|
| Advisor | 羅佩禎 |
|---|
| Publication year | 107 |
|---|
| Keyword | 腦電波=Electroencephalograph (EEG); 連續小波轉換=continuous wavelet transform (CWT); 機器學習=machine learning; 決策樹=decision tree (DT); 隨機森林=random forest (RF); 質心頻率=centroid frequency; 皮爾森相關係數=Pearson correlation coefficient (PCC); 禪定=Zen meditation |
|---|
| Abstract | This thesis is aimed to investigate the spatial-spectral properties of 30-channel Zen-meditation and resting EEG (electroencephalograph) based on classification of brain mapping of centroid frequency by decision tree (DT) and random forest (RF) models. Input data entry is the brain mapping of 30 centroid frequencies (BMFc) extracted from the CWT (continuous wavelet transform) coefficients of each channel. Based on the unsupervised learning scheme, DT mainly performs the matching of the input feature vector (brain mappings of the centroid frequency, BMFc) and the cluster center representing the quantitative features of the output cluster. Each DT with specified number of decision levels is trained by 1,000 BMFc’s. In addition, to optimize the clustering results, it is necessary to carefully select the implementation parameters and algorithm such as the decision-based input entry and cutoff point. From the clustering results, the major spatial-spectral features of Zen-meditation and resting EEG may be determined and compared. Random forest, a supervised learning scheme, is a classifier derived from ensemble learning. RF differs from DT in the aspect that RF is implemented with the randomness of the sampling and the decision on the ensemble output statistics. In other words, the random forest contains multiple decision trees constructed by random sampling of the data samples and makes the final decision by major voting among the decisions of all the DT trees. Finally, we compare the performance of the clustering results between decision tree (DT) and random forest (RF). Apparently, RF provides better clustering performance based on the measurement of precision (78% against 66%).
本論文中．我們利用機器學習，決策樹（DT）和隨機森林（RF）對禪定與休息狀態下腦電波加以分類，並試圖探索禪定與休息狀態下腦電波（30通道）質心頻率的空間特性。本研究採用非監督式分類方法及監督式分類方法，第一種方法是決策樹（DT），基於輸入特徵向量（brain mappings of 30 centroid frequencies, BMFc）和表示輸出簇（cluster）的定量特徵來匹配，DT由給30個輸入特徵（對於BMFc擷取的30通道）和決策層數建構。另外，為了優化分群的結果，有必要仔細的選擇參數與演算法，例如輸入的特徵，分割點。根據分群的結果，可以確定與比較禪定與休息腦電波質心頻率空間屬性的特徵。隨機森林（RF）是透過集成學習而得的分類器，與決策樹不同之處在於輸入取樣的隨機性與輸出的決策上．換句話說，隨機森林包含多棵隨機取樣建立的決策樹，並從所有決策樹通過投票法決定最後的輸出。最後，我們比較決策樹（DT）與隨機森林（RF）的分類效能，顯然RF的精確度是78%對上DT的66%，提供較好的分類效果。 |
|---|
| Table of contents | 摘要 i
Abstract ii
誌謝 iv
Content v
List of Figures viii
List of Tables xix
Chapter 1 1
Introduction 1
1-1 Background and Motivation 1
1-2 Aims of This Study 3
1-3 Scope of thesis 4
Chapter 2 5
Methods and Theories 5
2-1 Introduction of EEG 5
2-2 Continuous Wavelet Transform 7
2-3 Centroid Frequency analysis 9
2-4 Principal component analysis 9
2-5 Pearson correlation coefficient 12
2-6 Decision tree 13
2-6-1 CART algorithm 14
2-6-2 Feature selection 15
2-6-3 Split algorithm 16
2-6-4 Stopping criteria 17
2-7 Random forest 18
2-7-1 Bootstrap Aggregation 19
2-7-3 Out-of-bag 23
2-8 Evaluation of clustering performance 23
2-8-1 Cluster center 23
2-8-2 Intra-cluster distance 24
2-8-3 Intra-cluster PCC 24
2-8-4 Inter-cluster distance 25
2-8-5 Inter-cluster PCC 25
2-8-6 False-clustering 26
2-8-7 Precision 26
Chapter 3 28
Experiment and Signals Analysis 28
3-1 Experimental Setup and Procedure 28
3-2 Signal Analysis 29
3-2-1 Outline of the complete scheme 29
3-2-2 Datasets for DT and RF 31
3-2-3 Decision tree 32
3-2-4 Random forest 34
3-2-5 K-means clustering 40
3-3 Parameter Analysis for DT 42
3-3-1 Cutoff point (CPA) 43
3-3-2 Sort of similarity index 46
3-3-3 Number of decision levels (L) 52
3-4 Parameter Analysis for RF 62
3-4-1 Number of trees (T) 62
Chapter 4 65
Results and Discussion 65
4-1 Centroid frequency 65
4-1-1 Fc properties of Zen-meditation EEG 65
4-1-2 Fc properties of Resting EEG 72
4-2 Design of Input Dataset for DT construction 77
4-2-1 Results for Zen-meditation EEG 78
4-2-2 Results for Resting EEG 81
4-3 Assessment of Correct Clustering 84
4-3-1 Results for Zen-meditation EEG 84
4-3-2 Results for Resting EEG 85
4-4 BMFc classified by DT 86
4-4-1 Results for Zen-meditation EEG 86
4-4-2 Results for Resting EEG 117
4-4-3 Comparison between Zen-meditation EEG and Resting EEG 147
4-5 BMFc classification by RF 153
4-5-1 Results for Zen-meditation EEG 153
4-5-2 Results for Resting EEG 154
4-5-3 Comparison between Zen-meditation EEG and Resting EEG 155
Chapter 5 157
Conclusions and Discussions 157
5-1 Conclusions 157
5-2 Future work 171
Reference 172
Appendix A 180
Formal Zen-meditation Practice 180
Appendix B 181
B.1 Zen-meditation EEG and resting EEG in the 2nd segment 181
Zen-meditation EEG Type1 DT construction 181
Zen-meditation EEG Type2 DT construction 187
Zen-meditation EEG Type3 DT construction 191
Zen-meditation EEG Type4 DT construction 195
Resting EEG Type1 DT construction 199
Resting EEG Type2 DT construction 203
Resting EEG Type3 DT construction 207
Resting EEG Type4 DT construction 211
C.1 Zen-meditation EEG and resting EEG in the 1st segment 215
Zen-meditation EEG Type1 DT construction 215
Zen-meditation EEG Type2 DT construction 219
Resting EEG Type1 DT construction 223
Resting EEG Type2 DT construction 227
Resting EEG Type3 DT construction 231
Resting EEG Type4 DT construction 235
Appendix C 240 |
|---|
| Hits | 330 |
|---|
| Created date | 2022.10.13 |
|---|
| Modified date | 2023.02.17 |
|---|
 |
Best viewed with Chrome, Firefox, Safari(Mac) but not supported IE
|
|
|
