Passive Aggressive Ensemble for Online Portfolio Selection
Abstract
:1. Introduction
- The PAE framework is introduced to employ two distinct schemes to efficiently ensemble four different types of trend estimators with a lower time complexity of , where is the number of assets, is the number of trends, and is the total investment period.
- The PAE framework augments the performance of original trend estimators through reasonable evaluation and weighting.
- Extensive experiments are conducted on six real-world datasets to demonstrate that our algorithm not only outperforms competing strategies in terms of multiple evaluation criteria but also has promising scalability in terms of transaction costs.
2. Related Works and Problem Setting
2.1. Related Works
2.1.1. Trend Estimator
2.1.2. Ensemble Learning
2.2. Problem Setting
3. Methodology
3.1. Passive Aggressive Ensemble
3.2. Online Portfolio Selection with Multiple Trend Estimators
3.3. Solving Algorithm
Algorithm 1 PAE framework | ||
Input: | Given parameters , ,, , the trends and the actual price relatives in recent time window, the current weighting vector and the current portfolio . | |
Output: | The next portfolio . | |
1 Calculating the feasible portfolio by (9); | ||
2 Calculating the recent back-tested returns by (10) and by (11) to measure the performance of each trend; | ||
3 Defining the ensemble target by (12) and by (13); | ||
4 Setting parameter ; | ||
5 Updating weighting vector: ; | ||
6 Normalizing the next weighting vector by (22); | ||
7 Setting parameter by (24); | ||
8 Updating ; | ||
9 Normalizing the next portfolio by (25). |
3.4. Complexity Analysis
4. Experiments and Results
4.1. Data
4.2. Competing Portfolio Strategies
- BAH: the uniform Buy-And-Hold trading strategy. The strategy invests equally in assets at the onset and maintains this allocation throughout.
- OLMAR [9]: It takes the moving average to predict the future price. The parameters are set as follows: and .
- AICTR [12]: It combines three trends (SMA, EMA, PP) and market conditions through radial basis functions. The parameters are set as follows: , , and .
- SPOLC [17]: The short-term portfolio optimization with loss control strategy with the window size 𝑤 = 5 and the mixing parameter 𝛾 = 0.025.
- TPPT [11]: It uses adjustable historical windows and slope values for price prediction. The parameters are set as follows: , , and .
4.3. Evaluation Criteria
- Cumulative Wealth (CW). The CW serves as the principal metric for evaluating the investment performance of each portfolio selection algorithm. The CW is computed by (4).
- Annualized Percentage Yield (APY). The APY is a widely used metric for evaluating investment returns. It represents the average return of a strategy over the course of a year. APY is computed as follows:
- Sharpe Ratio (SR). In the realm of financial trading, it is often observed that higher returns are accompanied by elevated levels of risk. Thus, it is crucial for an investment algorithm to strike a balance between maximizing returns and managing risks. The SR serves as a widely utilized metric for evaluating risk-adjusted returns and is defined as follows:
- Calmar Ratio (CR) [34]. The CR is a comparison of the average annual compound return and the maximum drawdown (MDD) risk, which is widely adopted in fund management. The calculation formula is CR = APY/MDD, where .
4.4. Results
4.4.1. Parameter Setting
4.4.2. Ensemble Effectiveness
4.4.3. Comparison Studies
4.4.4. Transaction Costs
4.4.5. Running Time
5. Conclusions and Future Work
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
Appendix A
Appendix B
References
- Li, B.; Hoi, S.C.H. Online portfolio selection: A survey. ACM Comput. Surv. (CSUR) 2014, 46, 1–36. [Google Scholar] [CrossRef]
- Lai, Z.R.; Yang, H. A survey on gaps between mean-variance approach and exponential growth rate approach for portfolio optimization. ACM Comput. Surv. (CSUR) 2022, 55, 1–36. [Google Scholar] [CrossRef]
- Agarwal, A.; Hazan, E.; Kale, S.; Schapire, R.E. Algorithms for portfolio management based on the newton method. In Proceedings of the 23rd International Conference on Machine Learning, Pittsburgh, PA, USA, 25–29 June 2006; pp. 9–16. [Google Scholar]
- Györfi, L.; Udina, F.; Walk, H. Nonparametric nearest neighbor based empirical portfolio selection strategies. Stat. Risk Model. 2008, 26, 145–157. [Google Scholar] [CrossRef]
- Kahneman, D.; Tversky, A. Prospect theory: An analysis of decision under risk. In Handbook of the Fundamentals of Financial Decision Making: Part I; World Scientific: Singapore, 2013; pp. 99–127. [Google Scholar]
- Shiller, R.J. From efficient markets theory to behavioral finance. J. Econ. Perspect. 2003, 17, 83–104. [Google Scholar] [CrossRef]
- Li, B.; Zhao, P.; Hoi, S.C.; Gopalkrishnan, V. PAMR: Passive aggressive mean reversion strategy for portfolio selection. Mach. Learn. 2012, 87, 221–258. [Google Scholar] [CrossRef]
- Li, B.; Hoi, S.C.; Zhao, P.; Gopalkrishnan, V. Confidence weighted mean reversion strategy for online portfolio selection. ACM Trans. Knowl. Discov. Data (TKDD) 2013, 7, 1–38. [Google Scholar] [CrossRef]
- Li, B.; Hoi, S.C.; Sahoo, D.; Liu, Z.Y. Moving average reversion strategy for on-line portfolio selection. Artif. Intell. 2015, 222, 104–123. [Google Scholar] [CrossRef]
- Lai, Z.R.; Dai, D.Q.; Ren, C.X.; Huang, K.K. A peak price tracking-based learning system for portfolio selection. IEEE Trans. Neural Netw. Learn. Syst. 2017, 29, 2823–2832. [Google Scholar] [CrossRef]
- Dai, H.L.; Liang, C.X.; Dai, H.M.; Huang, C.Y.; Adnan, R.M. An online portfolio strategy based on trend promote price tracing ensemble learning algorithm. Knowl.-Based Syst. 2022, 239, 107957. [Google Scholar] [CrossRef]
- Lai, Z.R.; Dai, D.Q.; Ren, C.X.; Huang, K.K. Radial basis functions with adaptive input and composite trend representation for portfolio selection. IEEE Trans. Neural Netw. Learn. Syst. 2018, 29, 6214–6226. [Google Scholar] [CrossRef]
- Fern, A.; Givan, R. Online ensemble learning: An empirical study. Mach. Learn. 2003, 53, 71–109. [Google Scholar] [CrossRef]
- Von Krannichfeldt, L.; Wang, Y.; Hug, G. Online ensemble learning for load forecasting. IEEE Trans. Power Syst. 2020, 36, 545–548. [Google Scholar] [CrossRef]
- Helmbold, D.P.; Schapire, R.E.; Singer, Y.; Warmuth, M.K. On-line portfolio selection using multiplicative updates. Math. Financ. 1998, 8, 325–347. [Google Scholar] [CrossRef]
- Li, Y.; Zheng, X.; Chen, C.; Wang, J.; Xu, S. Exponential gradient with momentum for online portfolio selection. Expert Syst. Appl. 2022, 187, 115889. [Google Scholar] [CrossRef]
- Lai, Z.R.; Tan, L.; Wu, X.; Fang, L. Loss control with rank-one covariance estimate for short-term portfolio optimization. J. Mach. Learn. Res. 2020, 21, 3815–3851. [Google Scholar]
- Dai, H.L.; Huang, C.Y.; Dai, H.M.; Lai, F.T.; Lv, X.T.; Adnan, R.M. A novel adjusted learning algorithm for online portfolio selection using peak price tracking approach. Decis. Anal. J. 2023, 7, 100256. [Google Scholar] [CrossRef]
- Cai, X.; Ye, Z. Gaussian weighting reversion strategy for accurate online portfolio selection. IEEE Trans. Signal Process. 2019, 67, 5558–5570. [Google Scholar] [CrossRef]
- Cai, X. Vector autoregressive weighting reversion strategy for online portfolio selection. In Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence, Yokohama, Japan, 7–15 January 2021; pp. 4469–4475. [Google Scholar]
- Wu, B.; Lyu, B.; Gu, J. Weighted Multivariate Mean Reversion for Online Portfolio Selection. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases; Springer Nature: Cham, Switzerland, 2023; pp. 255–270. [Google Scholar]
- Guo, S.; Gu, J.W.; Fok, C.H.; Ching, W.K. Online portfolio selection with state-dependent price estimators and transaction costs. Eur. J. Oper. Res. 2023, 311, 333–353. [Google Scholar] [CrossRef]
- Guo, S.; Gu, J.W.; Ching, W.K. Adaptive online portfolio selection with transaction costs. Eur. J. Oper. Res. 2021, 295, 1074–1086. [Google Scholar] [CrossRef]
- Kumar, A.; Segev, A. Bayesian Ensembled Knowledge Extraction Strategy for Online Portfolio Selection. In Proceedings of the 2022 IEEE International Conference on Big Data (Big Data), Osaka, Japan, 17–20 December 2022; pp. 4148–4156. [Google Scholar]
- Duchi, J.; Shalev-Shwartz, S.; Singer, Y.; Chandra, T. Efficient projections onto the l1-ball for learning in high dimensions. In Proceedings of the 25th International Conference on Machine Learning, Helsinki, Finland, 5–9 July 2008; pp. 272–279. [Google Scholar]
- De Boer, P.T.; Kroese, D.P.; Mannor, S.; Rubinstein, R.Y. A tutorial on the cross-entropy method. Ann. Oper. Res. 2005, 134, 19–67. [Google Scholar] [CrossRef]
- Crammer, K.; Dekel, O.; Keshet, J.; Shalev-Shwartz, S.; Singer, Y.; Warmuth, M.K. Online Passive-Aggressive Algorithms. J. Mach. Learn. Res. 2006, 7, 551–585. [Google Scholar]
- Aldridge, I. High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems; John Wiley & Sons: Hoboken, NJ, USA, 2013; Volume 604. [Google Scholar]
- Borodin, A.; El-Yaniv, R.; Gogan, V. Can we learn to beat the best stock. In Proceedings of the Advances in Neural Information Processing Systems 16, Whistler, BC, Canada, 8–13 December 2003. [Google Scholar]
- Zhang, Y.; Lin, H.; Yang, X.; Long, W. Combining expert weights for online portfolio selection based on the gradient descent algorithm. Knowl.-Based Syst. 2021, 234, 107533. [Google Scholar] [CrossRef]
- Li, B.; Wang, J.; Huang, D.; Hoi, S.C. Transaction cost optimization for online portfolio selection. Quant. Financ. 2018, 18, 1411–1424. [Google Scholar] [CrossRef]
- Grinold, R.C.; Kahn, R.N. Active Portfolio Management; McGraw-Hill: New York, NY, USA, 2000. [Google Scholar]
- Treynor, J.L.; Black, F. How to use security analysis to improve portfolio selection. J. Bus. 1973, 46, 66–86. [Google Scholar] [CrossRef]
- Young, T.W. Calmar ratio: A smoother tool. Futures 1991, 20, 40. [Google Scholar]
- Lin, H.; Zhang, Y.; Yang, X. Online portfolio selection of integrating expert strategies based on mean reversion and trading volume. Expert Syst. Appl. 2024, 238, 121472. [Google Scholar] [CrossRef]
- Huang, D.; Zhou, J.; Li, B.; Hoi, S.C.; Zhou, S. Robust median reversion strategy for online portfolio selection. IEEE Trans. Knowl. Data Eng. 2016, 28, 2480–2493. [Google Scholar] [CrossRef]
Type | Strategies | Complexity |
---|---|---|
Single trend estimator | OLMAR | |
PPT | ||
Multiple trend estimators | AICTR | |
TPPT | ||
PAE |
Dataset | Region | Time | Days | Assets |
---|---|---|---|---|
NYSE(N) | US | 1 January 1985–30 June 2010 | 6431 | 23 |
TSE | CA | 1 January 1994–31 December 1998 | 1259 | 88 |
MSCI | Global | 1 April 2006–31 March 2010 | 1043 | 24 |
NYSE19 | US | 2 January 2015–4 September 2019 | 1167 | 47 |
ZZ28 | CN | 4 January 2000–1 April 2020 | 4905 | 28 |
ETF23 | CN | 1 February 2021–1 October 2023 | 647 | 23 |
Trend Estimator | NYSE(N) | MSCI | TSE | ZZ28 | NYSE19 | ETF23 |
---|---|---|---|---|---|---|
PP | 2.08 × 109 | 8.33 | 226.84 | 906.7 | 1.46 | 1.4 |
EMA | 4.64 × 108 | 23.6 | 680.83 | 283.58 | 2.46 | 1.69 |
SMA | 4.26 × 108 | 14.1 | 76.77 | 134.64 | 1.14 | 1.26 |
IP | 1.16 × 106 | 10.28 | 1.39 × 103 | 185.13 | 9.34 | 0.84 |
PAE-C | 6.83 × 108 | 23.63 | 706 | 348.31 | 2.51 | 1.29 |
PAE-R | 4.15 × 109 | 14.98 | 2.26 × 103 | 867.83 | 2.44 | 1.71 |
Dataset | Metrics | BAH | OLMAR | AICTR | SPOLC | TPPT | PAE-C | PAE-R |
---|---|---|---|---|---|---|---|---|
NYSE(N) | CW | 18.29 | 4.19 × 108 | 1.01 × 109 | 1.99 × 107 | 2.63 × 109 | 6.83 × 108 | 4.15 × 109 |
APY | 0.121 | 1.177 | 1.254 | 0.932 | 1.339 | 1.219 | 1.382 | |
SR | 0.046 | 0.104 | 0.106 | 0.105 | 0.108 | 0.105 | 0.113 | |
IR | −0.025 | 0.096 | 0.099 | 0.095 | 0.102 | 0.097 | 0.107 | |
CR | 0.225 | 1.28 | 1.374 | 1.082 | 1.561 | 1.298 | 1.629 | |
MSCI | CW | 0.89 | 14.5 | 12.38 | 7.34 | 10.81 | 23.63 | 14.98 |
APY | −0.027 | 0.908 | 0.837 | 0.618 | 0.776 | 1.147 | 0.923 | |
SR | 0.001 | 0.116 | 0.108 | 0.09 | 0.103 | 0.132 | 0.116 | |
IR | −0.036 | 0.169 | 0.158 | 0.132 | 0.155 | 0.193 | 0.17 | |
CR | −0.041 | 1.889 | 2.026 | 1.13 | 1.483 | 2.677 | 1.901 | |
TSE | CW | 1.56 | 57.79 | 544.47 | 277.1 | 265.10 | 706 | 2.26 × 103 |
APY | 0.093 | 1.252 | 2.529 | 2.083 | 2.055 | 2.717 | 3.692 | |
SR | 0.048 | 0.082 | 0.111 | 0.111 | 0.101 | 0.114 | 0.129 | |
IR | −0.002 | 0.078 | 0.108 | 0.107 | 0.098 | 0.111 | 0.127 | |
CR | 0.311 | 1.527 | 3.81 | 4.087 | 2.672 | 4.779 | 5.127 | |
ZZ28 | CW | 31.76 | 124.58 | 195.73 | 853.07 | 123.5 | 348.31 | 867.83 |
APY | 0.194 | 0.281 | 0.311 | 0.415 | 0.280 | 0.351 | 0.416 | |
SR | 0.048 | 0.05 | 0.053 | 0.068 | 0.050 | 0.057 | 0.063 | |
IR | 0.002 | 0.024 | 0.029 | 0.044 | 0.024 | 0.034 | 0.043 | |
CR | 0.335 | 0.389 | 0.436 | 0.711 | 0.430 | 0.474 | 0.548 | |
NYSE19 | CW | 1.37 | 0.98 | 1.79 | 2.45 | 1.56 | 2.51 | 2.44 |
APY | 0.07 | −0.005 | 0.133 | 0.214 | 0.101 | 0.22 | 0.213 | |
SR | 0.034 | 0.018 | 0.032 | 0.04 | 0.029 | 0.04 | 0.039 | |
IR | −0.022 | 0.009 | 0.024 | 0.033 | 0.021 | 0.033 | 0.032 | |
CR | 0.288 | −0.006 | 0.168 | 0.299 | 0.128 | 0.277 | 0.301 | |
ETF23 | CW | 0.87 | 1.19 | 1.28 | 0.84 | 1.07 | 1.29 | 1.71 |
APY | −0.053 | 0.07 | 0.102 | −0.064 | 0.027 | 0.104 | 0.231 | |
SR | −0.012 | 0.023 | 0.028 | −0.003 | 0.016 | 0.029 | 0.048 | |
IR | −0.005 | 0.036 | 0.043 | 0.005 | 0.028 | 0.043 | 0.067 | |
CR | −0.185 | 0.156 | 0.24 | −0.181 | 0.057 | 0.224 | 0.532 |
Statistics | NYSE(N) | MSCI | TSE | ZZ28 | NYSE19 | ETF23 | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
PAE-C | PAE-R | PAE-C | PAE-R | PAE-C | PAE-R | PAE-C | PAE-R | PAE-C | PAE-R | PAE-C | PAE-R | |
MER | 0.0032 | 0.0035 | 0.0033 | 0.0030 | 0.0066 | 0.0076 | 0.0007 | 0.0008 | 0.0011 | 0.0011 | 0.0010 | 0.0012 |
MER-market | 0.0006 | / | 0.0000 | / | 0.0004 | / | 0.0009 | / | 0.0004 | / | −0.0001 | / |
0.0030 | 0.0033 | 0.0033 | 0.0029 | 0.0061 | 0.0071 | 0.0007 | 0.0009 | 0.0009 | 0.0014 | 0.0008 | 0.0012 | |
1.3411 | 1.3561 | 1.2009 | 1.1946 | 2.2057 | 2.0434 | 1.0666 | 1.0718 | 1.7696 | 1.4239 | 1.1315 | 1.1495 | |
t-statistics | 7.3420 | 8.0863 | 6.3287 | 5.5531 | 3.6954 | 4.2855 | 2.2146 | 2.8055 | 1.3519 | 1.7481 | 1.1124 | 1.7443 |
p-value | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0002 | 0.0000 | 0.0268 | 0.0050 | 0.1769 | 0.0807 | 0.2664 | 0.0816 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Xie, K.; Yin, J.; Yu, H.; Fu, H.; Chu, Y. Passive Aggressive Ensemble for Online Portfolio Selection. Mathematics 2024, 12, 956. https://doi.org/10.3390/math12070956
Xie K, Yin J, Yu H, Fu H, Chu Y. Passive Aggressive Ensemble for Online Portfolio Selection. Mathematics. 2024; 12(7):956. https://doi.org/10.3390/math12070956
Chicago/Turabian StyleXie, Kailin, Jianfei Yin, Hengyong Yu, Hong Fu, and Ying Chu. 2024. "Passive Aggressive Ensemble for Online Portfolio Selection" Mathematics 12, no. 7: 956. https://doi.org/10.3390/math12070956
APA StyleXie, K., Yin, J., Yu, H., Fu, H., & Chu, Y. (2024). Passive Aggressive Ensemble for Online Portfolio Selection. Mathematics, 12(7), 956. https://doi.org/10.3390/math12070956