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

Classification of Zen-meditation and Resting EEG Spatial spectral properties by Random Forest

蔡宗諺(Tsung-Yen Tsai)

指導教授 ： 羅佩禎

國立交通大學/電機學院/電機與控制工程系所/碩士(2019年)

摘要

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.
關鍵字:腦電波,連續小波轉換,機器學習,決策樹,隨機森林,質心頻率,皮爾森相關係數,禪定.

關鍵字

腦電波 ； 連續小波轉換 ； 機器學習 ； 決策樹 ； 隨機森林 ； 質心頻率 ； 皮爾森相關係數 ； 禪定

並列摘要

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%).

Keywords: Electroencephalograph (EEG), continuous wavelet transform (CWT), machine learning, decision tree (DT), random forest (RF), centroid frequency, Pearson correlation coefficient (PCC), Zen meditation.

並列關鍵字

Electroencephalograph (EEG) ； continuous wavelet transform (CWT) ； machine learning ； decision tree (DT) ； random forest (RF) ； centroid frequency ； Pearson correlation coefficient (PCC) ； Zen meditation

參考文獻

[1] Huang, H-Y., and P-C. Lo. "EEG dynamics of experienced Zen meditation practitioners probed by complexity index and spectral measure." Journal of medical engineering & technology, vol. 33, no. 4, pp. 314-321, 2009.

[2] Lehmann, Dietrich, et al. "Reduced functional connectivity between cortical sources in five meditation traditions detected with lagged coherence using EEG tomography." Neuroimage, vol. 60, no. 2, pp. 1574-1586, 2012.

[3] Forman, Steven D., et al. "Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): use of a cluster‐size threshold." Magnetic Resonance in medicine, vol. 33, no. 5, pp. 636-647, 1995.

[4] Le Van Quyen, Michel, et al. "Nonlinear interdependencies of EEG signals in human intracranially recorded temporal lobe seizures." Brain Research, vol. 792, no. 1, pp. 24-40, 1998.

[5]Hämäläinen, Matti, et al. "Magnetoencephalography—theory, instrumentation, and applications to noninvasive studies of the working human brain." Reviews of modern Physics, vol. 65, no. 2, pp. 413-497, 1993.