在股票市場中，如何找出合適的買賣時機使得投資標的可獲利並降低風險是投資人重視的議題。為解決此問題，透過基本面、技術面或籌碼面指標組合而成的交易策略常用於決定標的買賣時機。由於技術面指標直接與股價相關且容易使用，故本論文主要著重於如何利用技術指標建立有效之交易策略。回顧文獻，相關研究議題包含；交易策略的制定、技術指標的參數最佳化、交易策略組合等。因現有技術所提供的交易策略組合有其限制，為了提昇交易策略組合的使用彈性與有效性，本論文首先定義群組交易策略組合最佳化問題，進而利用群組遺傳演算法設計兩群組交易策略組合最佳化方法。
在方法一中，每個群組交易策略組合透過群組、策略與資金權重三部分進行染色體編碼。每一染色體則透過利潤、風險、群組平衡與權重平衡四因子進行適合度評估。實驗透過上漲趨勢、盤整趨勢及下跌趨勢的資料集搭配兩交易策略集合與停損停利點進行有效性評估，實驗結果證實所提的方法可提供有效的群組交易策略組合。接著，因方法一實驗顯示使用停損停利點可提升利潤與降低風險，又設定合適的停損停利點是最佳化問題，故方法二在編碼上，除了使用群組、策略與資金權重，額外增加停損停利部分進行染色體編碼，之後透過群組遺傳演算法找出最佳之群組交易策略組合與其合適的停損停利點。實驗結果亦指出利用方法二所得的群組交易策略組合報酬明顯優於方法一。最後，本論文將所提的方法應用至群組股票投資組合上，實驗數據也顯示加入群組交易策略組合能有效降低股票投資組合風險且能的到更穩定的報酬。

In stock markets, how to determine an appropriate trading time for buying or selling stocks to make the return and risk of them being maximized and minimized is always an important issue for investors. The common way to deal with this problem is using trading strategies formed by fundamental, technical or chip analysis. Since there is a direct correlation between technical indicators and stock prices and technical indicators are much easy to use, hence, this thesis focus on how to establish efficient trading strategies by technical indicators. Literatures showed that there are lots of research topics, for example, including how to form trading strategies, parameter optimization for trading strategies, and trading strategy portfolio optimization. Because the trading strategy portfolios provided by existing approaches have limitations, to increase the flexibility and effectiveness of them, firstly, this thesis defines the group trading strategy portfolio optimization problem. Then, two group trading strategy portfolio optimization approaches are proposed using the grouping genetic algorithm.
In the first approach, a group trading strategy portfolio is encoded into a chromosome using three parts, the grouping, trading strategy and weight parts. The fitness function composes of four factors that are profit, risk, group balance and weight balance is utilized to assess the quality of a chromosome. Experiments were conducted on the uptrend, sideway trend and downtrend datasets with two sets of trading strategies and stop-loss and take-profit points to evaluate the effectiveness of the proposed approach. The experimental results show that the proposed approach can provide useful group trading strategy portfolio. Because the results also indicated that the first approach with stop-loss and take-profit points can increase return and reduce risk and to set stop-loss and take-profit points is an optimization problem, the second approach thus use not only grouping, trading strategy and weight but also stop-loss and take-profit part to encode a group trading strategy portfolio. Then, the grouping genetic algorithm is employed to optimize a group trading strategy portfolio and get its appropriate stop-loss and take-profit points. Experimental results reveal that the return of the second approach is better than the first approach. At last, the proposed approaches are applied on a group stock portfolio. The results show that stock portfolio with the group trading strategy can actually increase its ability to reduce risk and get more stable profit.

CHAPTER 1	INTRODUCTION	1
1.1 Motivation	1
1.2 Contributions	4
1.3 Reader’s Guide	5
CHAPTER 2	RELATED WORK	6
2.1 Review of the Trading Strategy Portfolio Optimization	6
2.2 Review of the Trading Strategy Parameter Optimization	7
2.3 Review of the Stock Portfolio Optimization with Trading Strategies	7
2.4 Literatures Related to Stop-Loss and Take-Profit Strategies	8
2.5 GGA and Grouping Problem	8
CHAPTER 3	PROBLEM DEFINITION	10
CHAPTER 4	AN EFFECTIVE APPROACH FOR OBTAINING A GROUP TRADING STRATEGY PORTFOLIO USING GROUPING GENETIC ALGORITHM	12
4.1 Motivation	12
4.2 Framework of the First Approach	13
4.3 The Four Components of the Proposed Approach	16
4.3.1 Encoding Scheme	16
4.3.2 Fitness Function and Reproduction	17
4.3.3 Genetic Operations	20
4.4 The Pseudo Code of the First Approach	22
4.5 An Example	23
CHAPTER 5	A SOPHISTICATED OPTIMIZATION ALGORITHM FOR OBTAINING A GROUP TRADING STRATEGY PORTFOLIO WITH APPROPRIATE STOP-LOSS AND TAKE-PROFIT POINTS	30
5.1 Motivation	30
5.2 Framework of the Second Approach	31
5.3 The Four Components of the Proposed Approach	34
5.3.1 Encoding Scheme	34
5.3.2 Fitness Function and Reproduction	37
5.3.3 Genetic Operations	39
5.4 The Pseudo Code of the Second Approach	41
5.5 An Example	43
CHAPTER 6	EXPERIMENTAL RESULTS	52
6.1 Data Descriptions	52
6.2 Experimental Results for Approach (I)	55
6.2.1 Experimental Results on the Three Datasets	56
6.2.1.1 Analysis of Uptrend Dataset	56
6.2.1.2 Analysis of Sideway Trend Dataset	61
6.2.1.3 Analysis of Downtrend Dataset	67
6.2.2 The Optimized GTSP	73
6.3 Experimental Results for Approach (II)	75
6.3.1 Analysis of Uptrend Dataset	75
6.3.1.1 Comparisons of the Proposed Approach and BHS	76
6.3.1.2 Comparisons of the Proposed Approach and the Previous Approach without SLTP Points	77
6.3.1.3 Comparisons of the Proposed Approach and the Previous Approach with Predefined SLTP Points	79
6.3.2 Analysis of Sideway Trend Dataset	82
6.3.2.1 Comparisons of the Proposed Approach and BHS	82
6.3.2.2 Comparisons of the Proposed Approach and the Previous Approach without SLTP Points	84
6.3.2.3 Comparisons of the Proposed Approach and the Previous Approach with Predefined SLTP Points	85
6.3.3 Analysis of Downtrend Dataset	88
6.3.3.1 Comparisons of the Proposed Approach and BHS	89
6.3.3.2 Comparisons of the Proposed Approach and the Previous Approach without SLTP Points	90
6.3.3.3 Comparisons of the Proposed Approach and the Previous Approach with Predefined SLTP Points	92
6.4 Application on a Group Stock Portfolio	95
CHAPTER 7	CONCLUSION AND FUTURE WORKS	98
REFERENCES	100
APPENDIX (I)	103
APPENDIX (II)	105

 
List of Figures
Figure 1. An example to explain the flexibility of a group trading strategy portfolio.	13
Figure 2. Data preprocessing procedure to form trading strategies.	14
Figure 3. Framework of the first approach.	15
Figure 4. Encoding schema of a GTSP.	16
Figure 5. A possible chromosome	17
Figure 6. Crossover on weight part.	21
Figure 7. The pseudo code of the first approach.	22
Figure 8. Framework of the second approach.	31
Figure 9. The Flowchart of trading strategies procedure.	33
Figure 10. Encoding schema of a GTSP and its SLTP points.	34
Figure 11. Chromosome representation	35
Figure 12. Crossover on weight part.	40
Figure 13. The pseudo code of the first approach.	42
Figure 14. Pseudo code of trading strategy procedure	43
Figure 15. The uptrend dataset.	53
Figure 16. The sideway trend dataset.	53
Figure 17. The downtrend of the dataset.	54


 
List of Tables
Table 1. Strategies and related information used in this example.	23
Table 2. The portfolio return of the ten chromosomes.	25
Table 3. Normalized MDD for every strategy.	26
Table 4. The risk of ten chromosomes.	26
Table 5. The group balances of all chromosomes.	27
Table 6. The weight balance of all chromosomes.	27
Table 7. The fitness values of all chromosomes.	27
Table 8.Trading strategy TS and stock price series	36
Table 9. Selected strategies and their related data.	44
Table 10. The portfolio return of the ten chromosomes.	46
Table 11. Normalized MDD for every trading strategy.	47
Table 12. The risk of ten chromosomes.	48
Table 13. The group balances of all chromosomes.	48
Table 14. The weight balance of all chromosomes.	49
Table 15. The fitness values of all chromosomes.	49
Table 16. The trading rules generated using the ten technical indicators.	55
Table 17. Returns of the GTSPs derived from one year training periods on one year testing periods on the uptrend dataset using the TOP15.	56
Table 18. Returns of the GTSPs derived from two year training periods on one year testing periods on the uptrend dataset using the TOP15.	57
Table 19. Returns of the GTSPs derived from three year training periods on one year testing periods on the uptrend dataset using the TOP15.	57
Table 20. Comparison returns of the derived GTSPs and BHS on the uptrend dataset using the TOP15.	58
Table 21. Returns of the GTSPs derived from one year training periods on one year testing periods on the uptrend dataset using TOP555.	59
Table 22. Returns of the GTSPs derived from two year training periods on one year testing periods on the uptrend dataset using TOP555.	60
Table 23. Returns of the GTSPs derived from three year training periods on one year testing periods on the uptrend dataset using TOP555.	60
Table 24. Comparison returns of the derived GTSPs and BHS on uptrend dataset using the TOP555.	61
Table 25. Returns of GTSPs trained from one year training periods on one year testing periods on the sideway trend dataset using TOP15.	62
Table 26. Returns of GTSPs trained from two year training periods on one year testing periods on the sideway trend dataset using TOP15.	62
Table 27. Returns of GTSPs trained from three year training periods on one year testing periods on the sideway trend dataset using TOP15.	63
Table 28. Comparison returns of the derived GTSPs and BHS on sideway trend dataset using the TOP15.	63
Table 29. Returns of GTSPs trained from one year training periods on one year testing periods on the sideway trend dataset using TOP555.	65
Table 30. Returns of GTSPs trained from two year training periods on one year testing periods on the sideway trend dataset using TOP555.	65
Table 31. Returns of GTSPs trained from three year training periods on one year testing periods on the sideway trend dataset using TOP555.	66
Table 32. Comparison returns of the derived GTSPs with BHS on sideway trend dataset using the TOP555.	66
Table 33. Returns of GTSPs trained from one year training periods on one year testing periods on the downtrend dataset using TOP15.	67
Table 34. Returns of GTSPs trained from two year training periods on one year testing periods on the downtrend dataset using TOP15.	68
Table 35. Returns of GTSPs trained from three year training periods on one year testing periods on the downtrend dataset using TOP15.	68
Table 36. Comparison returns of the derived GTSPs with BHS on downtrend dataset using the TOP15.	69
Table 37. Returns of GTSPs trained from one year training periods on one year testing periods on the downtrend dataset using TOP555.	70
Table 38. Returns of GTSPs trained from two year training periods on one year testing periods on the downtrend dataset using TOP555.	71
Table 39. Returns of GTSPs trained from three year training periods on one year testing periods on the downtrend dataset using TOP555.	71
Table 40. Comparison returns of the derived GTSPs with BHS on downtrend dataset using the TOP555.	72
Table 41. The optimized GTSPs by the proposed approach.	73
Table 42. Comparison of the proposed approach and BHS on the uptrend dataset.	76
Table 43. Comparison of the proposed approach and the previous approach without SLTP points on the uptrend dataset.	78
Table 44. Comparison of the proposed approach and the previous approach with SLTP points on the uptrend dataset.	79
Table 45. Comparison returns of the proposed approach and BHS on the sideway trend dataset.	82
Table 46. Comparison returns of the proposed approach and the previous approach without SLTP points on the sideway trend dataset.	84
Table 47. Comparison returns of the proposed approach and the previous approach with SLTP points on the sideway trend dataset.	86
Table 48. Comparison returns of the proposed approach and BHS on the downtrend dataset.	89
Table 49. Comparison returns of the previous approach and the previous approach without SLTP points on the downtrend dataset.	91
Table 50. Comparison returns of the proposed approach and the previous approach with SLTP points on the downtrend dataset.	93
Table 51. Comparison results of the GSPs with BHS and the proposed approach in terms of minimum returns.	95
Table 52. Comparison results of the GSPs with BHS and the proposed approach in terms of variance of returns.	96

[1]	Y. Chang and M. Lee, "Incorporating Markov decision process on genetic algorithms to formulate trading strategies for stock markets," Applied Soft Computing, Vol.52, No.10, pp. 1143–1153, 2016.
[2]	T. J. Chang, S. C. Yang and K. J. Chang, "Portfolio optimization problems in different risk measures using genetic algorithm," Expert Systems with Applications, Vol. 36, pp. 10529-10537, 2009.
[3]	Y. W. Chang-Chien and Y. L. Chen, "Mining associative classification rules with stock trading data–A GA-based method," Knowledge-Based Systems, Vol.23, No.6, pp. 605–614, 2010.
[4]	C. H. Chen, Y. H. Chen and M. E. Wu, "A GGA-based algorithm for group trading strategy portfolio optimization," The Multidisciplinary International Social Networks Conference, pp. 1-5, 2017.
[5]	J. Chen, J. Hou, S. Wu and Y. Chang-Chien, "Constructing investment strategy portfolios by combination genetic algorithms," Expert Systems with Applications, Vol. 36, No.2, pp. 3824-3828, 2009.
[6]	C. H. Chen, C. B. Lin and C. C. Chen, "Mining group stock portfolio by using grouping genetic algorithms," Evolutionary Computation (CEC), 2015 IEEE Congress on, pp. 738-743, 2015.
[7]	C. H. Chen, C. Y Lu, T. P Hong and J. H Su, "Using grouping genetic algorithm to mine diverse group stock portfolio," The IEEE Congress on Evolutionary Computation, pp. 4734 - 4738, 2016.
[8]	C. H. Chen, W. Y. Shen, T. P. Hong and J. H. Su, "An island-based algorithm for group stock portfolio optimization," Joint 17th World Congress of International Fuzzy Systems Association and 9th International Conference on Soft Computing and Intelligent Systems, pp. 1-5, 2017.
[9]	Y. H. Chou, S. Y. Kuo, C. Y. Chen and H. C. Chao, "A rule-based dynamic decision-making stock trading system based on quantum-inspired tabu search algorithm", IEEE Access, Vol.2, pp. 883-896, 2014.
[10]	Y. H. Chou, S. Y. Kuo and Y. T. Lo, "Portfolio optimization based on funds standardization and genetic algorithm," IEEE Access, Vol. 5, pp. 21885-21900, 2017.
[11]	E. Falkenauer, "A new representation and operators for genetic algorithms applied to grouping problems," Evolutionary Computation, Vol. 2, pp. 123-144, 1994.
[12]	E. Falkenauer, "A hybrid grouping genetic algorithm for bin packing," Journal of Heuristics, Vol. 2, pp. 5-30, 1996.
[13]	T. C. Fu, C. P. Chung and F. L. Chung, "Adopting genetic algorithms for technical analysis and portfolio management, "Computers and Mathematics with Applications, Vol.66,No.10, pp. 1743-1757, 2013.
[14]	D. E. Goldberg, "Genetic algorithms in search, optimization, and machine learning," Addison Wesley, 1989.
[15]	J. J. Grefenstette, "Optimization of control parameters for genetic algorithms," IEEE Transactions on System Man, and Cybernetics, Vol. 16, pp. 122-128, 1986.
[16]	F. Hassanzadeh, M. Collan and M. Modarres, "A practical approach to R&D portfolio selection using the fuzzy pay-off method," IEEE Transactions on Fuzzy Systems, Vol. 20, pp. 615-622, 2012.
[17]	L. R. Z. Hoklie, "Resolving multi objective stock portfolio optimization problem using genetic algorithm," International Conference on Computer and Automation Engineering, pp. 40-44, 2010.
[18]	J. H. Holland, "Adaptation in natural and artificial systems," University of Michigan Press, 1975.
[19]	K. M. Kaminski and A. W. Lo, "When do stop-loss rules stop losses?," Journal of Financial Markets, Vol.18, pp. 234-254, 2014.
[20]	Y. Kim and D. Enke, "Developing a rule change trading system for the futures market using rough set analysis," Expert Systems with Applications, Vol.59, pp. 165-173, 2016.
[21]	R. Kumar and S. Bhattacharya, "Cooperative search using agents for cardinality constrained portfolio selection problem," IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, Vol. 42, pp. 1510 - 1518, 2012.
[22]	S. Y. Kuo, C. Kuo and Y. H. Chou, "Dynamic stock trading system based on quantum-inspired tabu search algorithm," The IEEE Congress on Evolutionary Computation, pp. 1029-1036, 2013.
[23]	Y. Leu and T. I. Chiu, "An effective stock portfolio trading strategy using genetic algorithms and weighted fuzzy time series," The 16th North-East Asia Symposium on Nano, Information Technology and Reliability, pp. 70-75, 2011.
[24]	Y. J. Liu and W. G. Zhang, "Fuzzy portfolio optimization model under real constraints," Insurance: Mathematics and Economics, Vol. 53, No. 3, pp. 704-711, 2013.
[25]	A. W. Lo and A. Remorov, "Stop-loss strategies with serial correlation, regime switching, and transaction costs," Journal of Financial Markets, Vol.34, pp. 1-15, 2017.
[26]	K. Lwin, R. Qu and G. Kendall, "A learning-guided multi-objective evolutionary algorithm for constrained portfolio optimization," Applied Soft Computing, Vol. 24, pp. 757-772, 2014.
[27]	H. Markowitz, "Portfolio selection, "Journal of Finance," Vol. 7, No. 1, pp. 77-91, 1952.
[28]	H. M. Markowitz, "Harry Markowitz: Selected Works," World Scientific Publishing Company, 2009.
[29]	W. Nuij, V. Milea, F. Hogenboom, F. Frasincar and U. Kaymak "An automated framework for incorporating news into stock trading strategies," IEEE Transactions on Knowledge and Data Engineering, Vol. 26, No. 4, pp. 823-835, 2014.
[30]	X. Qin, H. Wang, F. Li, J. Chen, X. Zhou, X. Du, S. Wang, "Optimizing parameters of algorithm trading strategies using MapReduce," International Conference on Fuzzy Systems and Knowledge Discovery, pp. 2738-2741, 2012.
[31]	M. E. Wu, C. H. Wang and W. H. Chung, "Using trading mechanisms to investigate large futures data and their implications to market trends," Soft Computing, Vol. 21, No. 11, pp. 2821 - 2834, 2017.
[32]	H. Yao, Z. Li and D. Li, "Multi-period mean-variance portfolio selection with stochastic interest rate and uncontrollable liability," European Journal of Operational Research, Vol. 252, No. 3, pp. 837-851, 2016.