技術分析中的型態辨識問題在實務操作與學術理論中，一直以來都具有爭議性。而在技術分析的方法中，工程技術上最難的部分就是以一般化且自動化的方法來辨識歷史資料中的幾何型態，像是W底、M頭…等。
　　哼唱選歌是資訊工程領域的經典問題，在過去20年，已經成功的開發出許多人工智慧方法來解決這個問題。線性伸縮是效果最佳且最有效率的比對方法之ㄧ。我們應用這個方法到股價型態辨識上，使我們的演算法不僅能辨識傳統的股價型態，亦能辨識使用者自定義的股價型態。於文中將介紹我們的演算法、統計推論及實證結果。我們發現在型態形成後的報酬分布和隨機抽取的報酬分布，在台灣和美國NASDAQ的股票中，有統計上顯著的差異，但在中國A5O的成分股中，得到的統計結果並不顯著。
　　我們使用GPU進行平行運算，加速比對的過程。在實測上，能比在Matlab環境下快200倍；比使用C編譯快4倍。加速後的表現，使我們提出的型態辨識演算法，應用在即時環境下的可能性大幅提升。

關鍵字：技術分析、股價型態辨識、哼唱選歌、線性伸縮、GPU、平行運算

Technical analysis such as charting remains a disputing discipline between financial practice and academic finance. From the viewpoint of investment management in financial technology, the primary difficulty for adopting chart recognition is to identify various geometric shapes from historical price processes in a general and automated way. Query-by-singing/humming (QBSH in short) is a problem of music information retrieval in audio processing. Several artificial intelligence techniques have been developed successfully in computer science during last two decades. Linear scaling is well-known as one of the most efficient and robust methods for QBSH. We propose to apply this linear scaling method for identifying chart patterns, which can be traditional ones or user defined patterns.
Computational algorithms, statistical inference and empirical studies are implemented. Strong evidences on the discrepancies of stock return distributions before and after the presence of chart patterns are documented for Taiwan stock markets and NASDAQ in the US, but not the same case for A50 constitute stocks in China.
GPU parallel computing is further used to accelerate the process of chart recognition. In our experimental tests, speedup factors of 200 and 4 can be obtained under the programming environments Matlab and C, respectively. This numerical performance allows a potential usage of our proposed chart recognition technique for the real-time application.

Keywords: technical analysis, chart pattern recognition, Query-by-singing/humming, linear scaling method, GPU parallel computing.

Abstract i
Acknowledgement ii
Table of contents iii
Section 1: Introduction 1
Section 2: Traditional and Self-Defined Chart Patterns 5
2.1 Traditional Chart Patterns 6
2.2 Self-Defined Pattern 7
Section 3: Methodology: Applying The Linear Scaling Method of QBSH to Chart Pattern Recognition 7
3.1 QBSH for Music Information Retrieval and The Linear Scaling Method 8
3.2 Chart Pattern Recognition 10
3.2.1 Input Vectorization and Linear Scaling 10
3.2.2 Score Method 11
3.3 Comparison with other chart recognition methods: Correlation and Fourier Transform 12
3.3.1 Score by Correlation Coefficient 13
3.3.2 Score by L1 norm using Fourier Transform 13
3.3.3 Score by L1 norm 14
Section 4: Empirical Studies 15
4.1 Empirical Examples for Chart Recognition 16
4.2 Empirical Tests 18
4.2.1 Data 18
4.2.2 Goodness-of-Fit Test 19
4.3 Empirical Results 20
4.3.1 China Results 20
4.3.2 Taiwan Results 21
4.3.3 USA Results 22
4.4 GPU Acceleration 25
Section 5: Back Tests for Trading Strategy 27
5.1 Trading Strategy I 27
5.1.1 SSE 50 Index, China 28
5.1.2 Taiwan 50 Index, Taiwan 29
5.1.3 Dow Jones Index, U.S. A. 30
5.1.4 S&P 500 Index, U.S.A. 30
5.1.5 NASDAQ Capital Market, U.S.A. 31
5.2 Trading Strategy II 32
5.3 Trading Strategy III 36
Section 6: Conclusion 37
Reference 39
Appendix A 41
Appendix B 43

A. J. Ghias, D. C. Logan, and B. C. Smith, “Query by humming-musical information retrieval in an audio database,” in Proc. ACM Multimedia’95, San Francisco, pp. 216–221, 1995.
Andrew W. Lo, Harry Mamaysky, And Jiang Wang, 2000, Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation, Journal of Finance, 55 (4):1705-1765.
Bulkowski, Thomas N., 2000, “Encyclopedia of Chart Patterns,” John Wiley and Sons: New York.
Chung-Che Wang, Chieh-Hsing Chen, Chin-Yang Kuo, Li-Ting Chiu and Jyh-Shing Roger Jang , 2012, Accelerating QBSH On GPU: Optimization For Web Deployment, in Proc. of IEEE Conference on Acoustics, Speech and Signal Processing (ICASSP), p.477~p.480
Davidson, C. Chip and win: Banks expand use of GPUs, 25 March 2013, risk magazine.
C.L.Osler, P.H. Kevin Chang, 1995, Head and Shoulder: Not Just a Flaky Pattern, Federal Reserve Bank of New York Staff Reports, No. 4.
J.-S. R. Jang, H.-R. Lee, and M.-Y. Kao, “Content-based Music Retrieval Using Linear Scaling and Branch-and-Bound Tree search,” in Proc. of IEEE International Conference on Multimedia and Expo, August 2001, pp. 289-292
J.-S. R. Jang and M.-Y. Gao, “A query-by-Singing system based on dynamic programming,” in Proc. Int. Workshop Intell. Syst. Resolutions (8th Bellman Continuum), Hsinchu, Taiwan, R.O.C., Dec. 2000, pp. 85–89.
Ken-ichi Kainijo and Tetsuji Tanigawa, 1993, Stock Price Pattern Recognition: A Recurrent Neural Network Approach, Proceedings of the IEEE International Joint Conference on Neural Networks. San Diego. p.215–221
Roger Jang , 2012, Accelerating QBSH On GPU: Optimization For Web Deployment, in Proc. of IEEE Conference on Acoustics, Speech and Signal Processing (ICASSP), p.477 - p.480
R. J. McNab, L. A. Smith, I. H. Witten, C. L. Henderson, and S. J. Cunningham, “Toward the digital music library: Tune retrieval from acoustic input,” in Proc. ACM Digital Libraries, pp. 11–18, 1996.
Sang C. Suh, Dan Li, Jingmiao Gao, 2004, Novel Chart Pattern Recognition Approach: A Case Study On Cup With Handle, Proc of Artificial Neural Network in Engineering Conf, St. Louis, Missouri
S. Huang, L. Wang, S. Hu, H. Jiang, and B. Xu, “Query by humming via multiscale transportation distance in random query occurrence context,” Proc. ICME, 2008, pp. 1225-1228
W. Brock, J. Lakonishok, and B. LeBaron. 1992. Simple Technical Training Rules and the Stochastic Properties of Stock Returns, The Journal of Finance, 47(5), 1731–1764.
William J. O’Neil, How To Make Money In Stocks”, 3rd Edition, McGraw Hill
William Leigh, Noemi Paz, Russell Purvis, 2002, Market Timing: A Test Of A Charting Heuristic, Economics Letters, 77, 55-63
X. Wu, M. Li, J. Liu, J. Yang, and Y. Yan, “A top-down approach to melody match in pitch contour for query by humming,” Proc. International Conference of Chinese Spoken Language Processing, 2006, pp. 669-680
Y. Han, K. Yang, and G. Zhou. A New Anomaly: The Cross-Sectional Profitability of Technical Analysis. Journal of Financial and Quantitative Analysis. 48(5). pp. 1433-1461. October 2013.