在本論文中，我們將下載的克卜勒資料[12]，依據標準差大小分成三個模型，再以三個不同的機器學習模組(1D-CNN、2D-CNN-1、2D-CNN-2)對模型生成的資料庫進行訓練，進而希望透過這三個模組去尋找可能的系外行星。最後，我們發現標準差越大的模型其準確率較低，且在三個模組中以2D-CNN-2模組最佳。
在本論文中，我們發現在Kepler Q0的資料庫中，依據三種CNN的模組，並無找到任何系外行星。

In this thesis, we downloaded Kepler data [12] and divided into three models based on the magnitude of standard deviations. We used three different machine learning modules such as 1D-CNN, 2D-CNN-1 and 2D-CNN-2 to train the database generated by the model. It hopes to find possible exoplanets through these three modules. Finally, we found that the model with larger standard deviation has lower accuracy. Among the three modules, we found that the 2D-CNN-2 is the best module.
In this thesis, we used three kinds of CNN modules, and found that there is no any exoplanet in Kepler Q0 data set.

致謝...................................I
摘要..................................II
Abstract.............................III
第一章 簡介.............................1
第二章 資料處理.........................3
2.1 原始資料............................3
2.2 資料標準化..........................4
2.3 資料刪除離群值......................4
2.4 時間間隔模組........................5
第三章 系外行星理論模型..................6
3.1 系外行星模型參數.....................6
3.2 資料分類............................7
3.3 資料折疊............................8
3.4 資料內插...........................10
3.5 資料折疊週期±2分鐘(P±21440).........10
第四章 卷積神經網路模型..................12
4.1 卷積層(Convolution Layer)..........12
4.1.1 一維卷積層簡介....................12
4.1.2 二維卷積層簡介....................14
4.2 池化層(Pooling Layer)..............15
4.3 丟棄層(Dropout Layer)..............16
4.4 全連接層(Fully Connected Layer)....17
第五章 系外行星的CNN模組介紹.............18
5.1 1D-CNN模型與參數選取................18
5.1.1 系外行星1D-CNN模組................19
5.2 2D-CNN模型與參數選取 ...............20
5.2.1 系外行星2D-CNN-1模組..............21
5.2.2 系外行星2D-CNN-2模組..............22
5.3 交叉驗證(Cross-Validation, CV)......23
第六章 測試結果..........................25
6.1 模型A...............................26
6.2 模型B...............................30
6.3 模型C...............................35
6.4 尋找週期............................39
第七章 結論..............................40
參考文獻.................................41
附錄.....................................42
A、名詞對照表............................42
B、模型編號表............................44
C、程式碼................................47
C.1 1D-CNN..............................47
C.2 2D-CNN-1............................54
C.3 2D-CNN-2............................60

[1] Chintarungruangchai, P., & Jiang, G. (2019). Detecting Exoplanet Transits through Machine-learning Techniques with Convolutional Neural Networks. Publications of the Astronomical Society of the Pacific, 113(1000), 064502.
[2] Kohavi, R. (1995). A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection. The International Joint Conference on Artificial Intelligence, 1137-1143.
[3] Mandel, K., & Agol, E. (2002). Analytic light curves for planetary transit searches. The Astrophysical Journal, 580(2), L171.
[4] Nikhil Ketkar, Deep Learning with Python, Springer,2017
[5] Shallue, C. J. & Vanderburg, A. (2018). Identifying Exoplanets with Deep Learning: A Five-planet Resonant Chain around Kepler-80 and an Eighth Planet around Kepler-90. The Astronomical Journal, 155(2), 94.
[6] Srivastava, N., et al. (2014). Dropout: A Simple Way to Prevent Neural Networks from Overfitting. The Journal of Machine Learning Research, 15, 1929-1958.
[7] Yeh, C., & Jiang, G. (2020). Searching for Possible Exoplanet Transits from BRITE Data through a Machine Learning Technique. Publications of the Astronomical Society of the Pacific,133(1019), 014401.
[8] 陳輝樺，「找尋太陽系外行星的方法」，科博館訊，306，2013，7.
[9] 胡佳玲，「系外行星」，臺北星空，92，2019，8-12.
[10] 郭芷綺，「以機器學習法搜尋戲外行星的研究」，國立清華大學，碩士，109
[11] https://www.nasa.gov/kepler/missiontimeline
[12] https://exoplanetarchive.ipac.caltech.edu/bulk_data_download/
[13] https://www.natgeomedia.com/science/video/content-7744.html
[14] https://exoplanets.nasa.gov/alien-worlds/ways-to-find-a-planet/
[15] https://www.researchgate.net/figure/Dropout-neural-network-model-a-is-a-standard-neural-network-b-is-the-same-network_fig3_309206911