人們在進行金融交易時，經常會依靠直覺來進行思考。舉例而言，交易員經常一次接收多個螢幕的資訊並快速地做出決定，但往往詢問他們所得到的決策理由大多是依靠長年累積而來的「盤感」(domain knowledge)。為了去模擬出人們腦中的高複雜度模型，我們決定使用深度學習模型中的卷積神經網絡模型 (Convolutional Neural Networks, CNNs) 來進行此任務。卷積神經網絡模型在圖像識別領域上非常突出，這點如同利用電腦模擬人們的眼睛去進行看盤的行為一樣，這也是我們選擇卷積神經網絡模型而不採行常見的量化分析模型去建模的主要原因。

我們使用的資料是由幾何布朗運動過程 (Geometric Brownian Motion process) 所模擬出來的日幣走勢。我們會先把外匯的量化資料預處理成圖片，這些圖片包含了外匯價格、五日移動平均線 (Moving Average 5)、十日移動平均線 (Moving Average 10)，以及二十日移動平均線 (Moving Average 20) 等等的資訊，並且以此作為輸入層 (input layer) 輸入卷積神經網絡模型之中，通過不斷試驗不同的結構與模型參數去進行實驗，最後再以各分類的準確度作為衡量模型的標準。我們的最終目標是在找出能夠極大化分類準確度的模型，同時在驗證卷積神經網路是否能夠複製人們的交易策略。

Deep learning is an effective approach to solve image recognition problems.
People like to think intuitively from the trading chart. This study used the characteristics of deep learning to train computers how to imitate people''s thinking from the trading chart.

We have three steps as follows:

1. Before training, we need to pre-process our input data from quantitative data to images.
2. We use Convolutional-Neural-Network (CNN), which is a kind of the deep learning, to train our trading model.
3. We evaluate the model performance by the accuracy of classification.

With this approach, a trading model is obtained to help make trading strategies. The main application is designed to help clients automatically obtain personalized trading strategies.

Abstract iii
List of Figures ix
Chapter 1: Introduction ... 1
1.1 Motivation ... 1
1.2 Related Works ... 2
1.3 Problem Definition ... 2
1.4 Structures of Dissertation ... 3
Chapter 2: Literature Review ... 5
Chapter 3: Mathematics ... 7
3.1 Geometric Brownian Motion (GBM) ... 7
3.2 Convolutional Neural Networks (CNNs) ... 8
3.2.1 History of the CNNs model ... 8
3.2.2 Introduce to CNNs model ... 9
3.2.3 Architectures of our CNNs: some architectures we have tried ... 11
3.2.4 Architectures of our CNNs model: AlexNet ... 14
Chapter 4: Workflow ... 15
4.1 Introduction to our workflow ...15
4.2 Workflow 1 ... 15
4.3 Workflow 2 ... 16
Chapter 5: Discussion ... 19
5.1 Workflow 1 ... 19
5.1.1 Architecture 1 ... 21
5.1.2 Architecture 2 ... 21
5.1.3 Architecture 3 ... 22
5.2 Workflow 2 ... 23
5.2.1 experiments 1 ... 24
5.2.2 experiments 2 ... 25
5.2.3 experiments 3 ... 26
Chapter 6: Conclusions ... 37
Bibliography ... 39
Index

[1] S. Browne. Optimal investment policies for a firm with a random risk process: Ex-ponential utility and minimizing the probability of ruin. Mathematics of Operations Research, 20(4):937–958, 1995.

[2] L. Di Persio and O. Honchar. Artificial neural networks approach to the forecast of stock market price movements.

[3] K. Fukushima and S. Miyake. Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition. In Competition and cooperation in
neural nets, pages 267–285. Springer, 1982.

[4] S. E. Shreve. Stochastic Calculus for Finance II: Continuous-Time Models. Springer, New York, 2004.

[5] H. Wang, B. Raj, and E. P. Xing. On the origin of deep learning. arXiv preprint arXiv:1702.07800, 2017.