在多變的金融市場中，多樣的技術分析和基本指標被用來產生交易策略並模組化用，旨在用來確定買賣股票時能夠做出適合的交易決策。在文獻中曾提出了群組交易策略組合最佳化方法來最佳化一個群組交易策略組合並制定最佳交易決策且其屬於單目標最佳化問題。然而，在現實的情況中，投資者必須同時考量多個目標來制定交易決策。故本篇論文提出一個多目標為基礎的演算法來找出柏拉圖解並提供給使用者建置更有用的交易策略。其中，每一柏拉圖解即為一群組交易策略組合。 為了最佳化一群組交易策略組合，根據所選擇的技術指標，演算法先產生候選交易策略。然後，再利用所設計之排名函數選出交易策略子集合。根據此交易策略子集合，所提的方法接著產生初始母體與用來記錄柏拉圖解的空集合。在編碼方式上，每個可能的群組交易策略組合由三部分來表達，分別為：群組、權重與交易策略。在演化過程中，兩個目標函數被用來評估每個染色體的適合度並找尋柏拉圖解。第一個目標函數將用來評估染色體的獲利與風險。第二個目標函數則用來衡量染色體的群組與權重平衡度。遺傳運算則接續用來產生新的染色體，包含：交配、突變與反轉運算。在實驗部分，我們透過三個不同趨勢的股價資料來展示所提的方法的有效性，分別為：上升趨勢、盤整與下降趨勢。

A variety of technical analyses techniques and fundamental indicators have been used to form trading strategies and modeled to determine the appropriate trading decisions for when to sell or buy stocks at unstable challenging financial market. A group trading strategy optimization portfolio algorithm was presented in the literature to find out an optimal group trading strategy portfolio to make trading decisions, and it belongs to the single objective optimization problem. However, in the real situation, traders have confronted to make decision by considering multiobjective goals. Hence, this thesis proposes a MOGA-based algorithm to find a set of Pareto solutions for investors to make more useful trading plans, and each solution is a group trading strategy portfolio. To optimize a GTSP, the candidate trading strategies are first produced according to the chosen technical indices. Then, a subset of the candidate trading strategies is selected using the determined ranking functions. Based on the subset of the trading strategies, the population is initialized as determined chromosome, and non-dominated set is initialized as empty. In the encoding scheme, the grouping, weighting and trading strategy parts are utilized to represent a possible GTSP.  The two objective functions are used to evaluate the fitness values of chromosomes to discover non-dominated solutions. The first objective function is used to evaluate the return and risk of a GTSP in the chromosome. The second objective function is utilized to reveals the grouping and weight balances of the trading strategy groups. The genetic operators, including crossover, mutation, and inversion are executed on the population to generate new offspring. In the experiment, the proposed algorithm is evaluated on three datasets with different trends, namely uptrend, sideway trend and downtrend, to show the effectiveness of the proposed approach.

Contents
CHAPTER 1	1
INTRODUCTION	1
1.1	Problem Definition	1
1.2	Contributions	2
1.3	Reader’s Guide	3
CHAPTER 2	4
LITERATURE REVIEWS	4
2.1	Review of Stock Portfolio Optimization with Trading Strategies	4
2.2	Review of Trading Strategies Optimization	6
2.3	Review of Stop-Loss and Take-Profit Strategies	9
CHAPTER 3	10
DEFINITION OF PROBLEM AND FRAMEWORK OF THE PROPOSED APPROACH	10
3.1	Definition of Problem	10
3.2	Framework of the Proposed Approach	11
CHAPTER 4	14
COMPONENTS OF PROPOSED APPROACH	14
4.1	Encoding Schema	14
4.2	The Two Objective Functions	16
4.3	The Rank-based Fitness assignment	19
4.4	Genetic Operations	22
CHAPTER 5	25
PROPOSED ALGORITHM	25
5.1	Pseudo Code of the Proposed Approach	25
5.2	Steps of the Proposed Algorithm	27
5.3	An Example	29
CHAPTER 6	41
EXPERIMENTAL RESULTS	41
6.1	Data Descriptions	41
6.2	Experimental Result Evaluation on Uptrend Dataset.	45
6.3	Experimental Result Evaluation on Sideway Trend Dataset	48
6.4	Experimental Result Evaluation on Downtrend Dataset	50
6.5	Evaluation on the Derived Non-Dominated Solutions	53
CHAPTER 7	55
CONCLUSIONS AND FUTURE WORK	55
REFERENCES	56
APPENDIX	60

List of figures
Figure 1. Framework of the proposed approach.	12
Figure 2. Encoding schema for a GTSP.	14
Figure 3. An initial chromosome.	15
Figure 4. The ranking results of the 15 chromosomes.	19
Figure 5. The results of assign fitness of the 15 chromosomes.	21
Figure 6. The average fitness values of the 15 chromosomes.	22
Figure 7. The uptrend datasets.	42
Figure 8. The sideway trend datasets.	42
Figure 9. The downtrend datasets.	43
Figure 10. Non-dominated Pareto front of uptrend dataset.	46
Figure 11. Non-dominated Pareto front of sideway trend dataset.	48
Figure 12. Non-dominated Pareto front of downtrend dataset.	51


List of Tables
Table 1. The crossover operator on the weight part.	23
Table 2. The multiobjective genetic algorithm.	25
Table 3. Selected strategies and related information used in the example.	29
Table 4. The objective function 1 of the ten chromosomes.	32
Table 5. The portfolio returns of the ten chromosomes.	32
Table 6. Normalized MDD for every strategy.	33
Table 7. The risk of ten chromosomes.	34
Table 8. The objective function 2 of the ten chromosomes.	34
Table 9. The group balances of all chromosomes.	35
Table 10. The weight balance of all chromosomes.	35
Table 11. The multiobjective fitness values of all chromosomes.	36
Table 12. The ranking results of all the 15 chromosomes.	37
Table 13. The fitness values of all the 15 chromosomes.	38
Table 14. The resulting average fitness values of the 15 chromosomes.	38
Table 15. The trading rules generated using the ten technical indicators.	44
Table 16. Comparison of ROI between GGA-GTSP, MOGA-GTSP and BHS at uptrend datasets.	46
Table 17. Comparison of ROI between GGA-GTSP, MOGA-GTSP and BHS at sideway trend dataset.	49
Table 18. Comparison of ROI between GGA-GTSP, MOGA-GTSP and BHS at downtrend dataset.	51
Table 19. The comparison of optimized MOGA-GTSP, GGA-GTSP and BHS on the three trend datasets.	53

REFERENCES

[1]	 Kim Youngmin, Enke David, “Developing a rule change trading system for the futures market using rough set analysis,” Expert Systems with Applications, Vol. 59, pp. 165-173, 2016.
[2]	Y. Kim, W. Ahn, K.J. Oh and D. Enke, “An intelligent hybrid trading system for discovering trading rules for the futures market using rough sets and genetic algorithms,” Applied Soft Computing Journal, Vol.55, pp.127-140, 2017.
[3]	Y.Chen and X.Wang, “A hybrid stock trading system using genetic network programming and mean conditional value-at-risk.” European Journal of Operational Research, Vol. 240, pp.861-871, 2015.
[4]	Bahar. H.,  Zarandi, M.,  Esfahanipour, A. “A hybrid expert system for generating stock trading signals,” International Journal of Computer, Electrical, Automation, Control and Information Engineering, Vol. 115, pp. 1295 – 1300, 2016.
[5]	A. B. Prasetijo, T. A. Saputro, I. P. Windasari and Y. E. Windarto, "Buy/sell signal detection in stock trading with bollinger bands and parabolic SAR: With web application for proofing trading strategy,"  4th International Conference on Information Technology, Computer, and Electrical Engineering, pp. 41-44, 2017.
[6]	L. Wang, "Dynamical models of stock prices based on technical trading rules—part III: Application to Hong Kong stocks," IEEE Transactions on Fuzzy Systems, Vol. 23, no. 5, pp. 1680-1697, 2015
[7]	Y. Ma and R. Han, "Research on stock trading strategy based on deep neural network," 18th International Conference on Control, Automation and Systems, pp. 92-96, 2018.
[8]	R. Ruiz-Cruz and A. D. Diaz-Gonzalez, "Investment portfolio trading based on Markov chain and fuzzy logic," IEEE Latin American Conference on Computational Intelligence, pp. 1-6, 2018.
[9]	C. H. Chen, Y. H. Chen, J. C. W. Lin and M. E. Wu, "An effective approach for obtaining a group trading strategy portfolio using grouping genetic algorithm," IEEE Access, Vol. 7, pp. 7313-7325, 2019.
[10]	J. Pinto, R. F. Neves and N. Horta, “Multi-objective optimization of investment strategies based on evolutionary computation techniques, in volatile environments,” Proceedings of the 16th International Conference on Enterprise Information System, pp. 480-488, 2014. 
[11]	D. J. Bodas Sagi, F. J. Soltero, J. I. Hidalgo, P. Fernández and F. Fernandez, "A technique for the optimization of the parameters of technical indicators with Multi-Objective Evolutionary Algorithms," IEEE Congress on Evolutionary Computation, pp. 1-8, 2012.
[12]	Y. H. Chou, S. Y. Kuo and C. Kuo, "A dynamic stock trading system based on a multi-objective quantum-inspired tabu search algorithm," The IEEE International Conference on Systems, Man, and Cybernetics, pp. 112-119, 2014.
[13]	D. Lohpetch and D. Corne, "Multi-objective algorithms for financial trading: multi-objective out-trades single-objective," IEEE Congress of Evolutionary Computation, pp. 192-199, 2011.
[14]	A.C Briza and P.C Naval, “Stock trading system based on the multi-objective particle swarm optimization of technical indicators on end-of-day market data,” Applied Soft Computing, 2010.
[15]	M. Rajabi and H. Khaloozadeh, "Investigation and comparison of the performance of multi-objective evolutionary algorithms based on decomposition and dominance in portfolio optimization," Electrical Engineering Iranian Conference, pp. 923-929, 2018.
[16]	T. Murata and H. Ishibuchi, "MOGA: multi-objective genetic algorithms," IEEE International Conference on Evolutionary Computation, pp. 289-294, 1995. 
[17]	1.2 R. de Almeida, G. Reynoso-Meza and M. T. A. Steiner, "Multi-objective optimization approach to stock market technical indicators," The IEEE Congress on Evolutionary Computation, pp. 3670-3677, 2016.
[18]	D.A. Silva, F.N. Rui and N. Horta, “Portfolio optimization using fundamental indicators based on multi-objective EA” pp.39-56, 2016. 
[19]	Hoklie and L. R. Zuhal, "Resolving multi objective stock portfolio optimization problem using genetic algorithm," The 2nd International Conference on Computer and Automation Engineering, pp. 40-44, 2010.
[20]	R. Drezewski, K. Doroz, “An agent-based co-evolutionary multi-objective algorithm for portfolio optimization,” Symmetry, 2017.
[21]	1.2 K. Lwin, R. Qu, G. Kendall, “A learning-guided multi-objective evolutionary algorithm for constrained portfolio optimization,” Applied Soft Computing, Vol. 24, pp.757-772, 2014.
[22]	S. Babaei, M.M Sepehri, E. Babaei, “Multi-objective portfolio optimization considering the dependence structure of asset returns,” European Journal of Operational Research, 2015. 
[23]	R. Ramadhiani, M. Yan, G. F. Hertono and B. D. Handari, "Implementation of e-new local search based multi-objective optimization algorithm and multi-objective co-variance based artificial bee colony algorithm in stocks portfolio optimization problem," 2nd International Conference on Informatics and Computational Sciences, pp. 1-6, 2018.
[24]	W. Si, J. Li, P. Ding and R. Rao, "A multi-objective deep reinforcement learning approach for stock index future’s intraday trading," 10th International Symposium on Computational Intelligence and Design, pp. 431-436, 2017.
[25]	G. A. V. Pai, "Multi-objective metaheuristics for managing futures portfolio risk," IEEE Symposium Series on Computational Intelligence, pp. 1204-1211, 2018.
[26]	Zh.Huiming and J. Watada, “A fuzzy index tracking multi-objective approach to stock data analytics,” 4th International Conference on Computer and Information Sciences, 2018.
[27]	Z. Liu, Z. Liu, Y. Song, Z. Gong and H. Chen, "Predicting stock trend using multi-objective diversified echo state network," Seventh International Conference on Information Science and Technology, pp. 181-186, 2017.
[28]	C. M. Fonseca and P. J. Fleming, "Genetic algorithms for multiobjective optimization: Formulation, discussion and generalization," The International Confidence on Genetic Algorithms, pp. 416-423, 1993.
[29]	I. Ucar, A. M. Ozbayoglu and M. Ucar, "Developing a two level options trading strategy based on option pair optimization of spread strategies with evolutionary algorithms," IEEE Congress on Evolutionary Computation, pp. 2526-2531, 2015.
[30]	R. Ji, M. A. Lejeune and S. Y. Prasad, "Dynamic portfolio optimization with risk-aversion adjustment utilizing technical indicators," 20th International Conference on Information Fusion, pp. 1-8, 2017.
[31]	K. M. Kaminski and A. W. Lo, "When do stop-loss rules stop losses?" Journal of Financial          Markets, Vol.18, pp.234-254, 2014
[32]	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