Stock portfolio optimization using deep reinforcement learning based on convolutional neural network model

Stock portfolio optimization using deep reinforcement learning based on convolutional neural network model

Subject Areas : Financial Knowledge of Securities Analysis

Sanaz Dalili ¹ , Yasser Rezaei Pitenoei ² , mohammadtaghi kabiri ³ , مهدی صفری گرایلی ⁴ , meysam arabzadeh ⁵

1 - PhD student, Department of Financial Engineering, Kashan Branch, Islamic Azad University, Kashan, Iran
2 - Associate Professor , Department of Accounting, University of Guilan, Rasht, Iran. (Responsible author).
3 - Assistant Professor, Department of Accounting, Faculty of Administrative and Economic Sciences, Arak University, Arak, Iran
4 - Associate Professor Department of Accounting,Bandargaz Branch, Islamic Azad University, Bandargaz, Iran
5 - Assistant Professor, Department of Accounting, Kashan Branch, Islamic Azad University, Kashan, Iran.

Received: 2024-04-15 Accepted : 2024-05-21 Published : 2025-07-22

Keywords: Optimization, stock portfolio, deep reinforcement learning, convolutional neural network.,

Abstract :

Classic portfolio optimization models can be used when capital market constraints and portfolio formation criteria are low. But considering the realities of the capital market and the weakness of classical models in discovering complex and non-linear relationships, it can be understood that the problem is more complex and beyond the classical models. In solving such complex problems, interdisciplinary sciences, computers and models based on machine learning have come to the aid of humans and have had a successful approach in solving these problems. Recently, deep learning techniques, such as recurrent neural network, have been designed for portfolio optimization. In this research, a deep reinforcement learning convolutional neural network model is proposed for stock portfolio optimization based on historical data. In this sense, the daily returns of 20 companies admitted to the Tehran Stock Exchange, which had high liquidity in the period from 1392 to 1401, have been used. The results obtained from this research show that based on the evaluation of the portfolio return, risk index and Sharpe index and that in all criteria the results of the deep learning convolution neural network model portfolio were optimal, it can be said that the neural network model Deep learning convolution has a good ability to form an optimal stock portfolio.

References:

تهرانی، رضا، فلاح تفتی، سیما و آصفی، سپهر.(1397). بهینه سازی سبد سهام به کمک الگوریتم فراابتکاری دسته های میگو با استفاده از معیارهای مختلف از ریسک در بورس اوراق بهادار تهران، فصلنامه تحقیقات مالی، 20(4).87-101.
حیدری، محمدسعید و ولیدی، جواد و ابراهیمی، سیدبابک.(1400). بهینه سازی سبد سهام مبتنی بر مدل برنامه ریزی امکانی استوار با استفاده از الگوریتم های ژنتیک و جهش قورباغه مخلوط شده، مهندسی مالی و مدیریت اوراق بهادار، 12(47).112-127.
فتاحی نافچی، حسن، عرب صالحی، مهدی، اسماعیلیان، مجید. (1398). انتخاب سبد بهینه سهام با بکارگیری اطلاعات حسابداری، اطلاعات مبتنی برارزش و اطلاعات کارت ارزیابی متوازن -مورد مطالعه: شرکت‌های پذیرفته شده در بورس اوراق بهادار تهران. پیشرفت‌های حسابداری، 11(2): 285-320.
کرکه آبادی، مهدیه و کلهری، معصومه.(1399).شناسایی و رتبه بندی عوامل موثر بر انتخاب بهینه سبد سهام با استفاده از روش AHP فازی (مطالعه موردی: بورس اوراق بهادار)،هشتمین همایش ملی پژوهش های مدیریت و علوم انسانی در ایران،تهران.
نیکو، سیده فرخ، شمس، شهاب الدین، صیقلی، محسن. (1399). مدل‌سازی انتخاب سبد بهینه سهام بر مبنای ارزیابی ریسک و رویکرد مالی رفتاری (حسابداری ذهنی) در بورس اوراق بهادار تهران. نشریه چشم انداز مدیریت مالی، 10(31): 75-101.
یزدانی خداشهری محمدباقر، نسل موسوی سید حسین، حسینی شیروانی میرسعید.(1400)/ انتخاب بهینه سبد سهام با استفاده از الگوریتم ترکیبی فرا ابتکاری. تحقیق در عملیات در کاربردهای آن. ۱۸ (۱) :۱۰۱-۱۲۴.
Anagnostopoulos, K.P., Mamanis, G. (2010). A portfolio optimization model with three objectives and discrete variables. Computers & Operations Research, 37 (7), 1285-1297.
Arel, I. Rose, D.C. Karnowski, T.p. (2010). "Deep Machine Learning - A New Frontier in Artificial Intelligence Research [Research Frontier]". IEEE Computational Intelligence Magazine, 5: 13-.81
Bertsimas, D., Shioda, R. (2009). Algorithm for cardinality-constrained quadratic optimization. Computational Optimization and Applications, 43(1), 1–22.
Carandente, Vincenzo & Sperlì, Giancarlo. (2024). Harnessing Cognitively Inspired Predictive Models to Improve Investment Decision-Making. Cognitive Computation18(6). 1-16.
Chang, T. J., Meade, N., Beasley, J. E. & Sharaiha, Y. M. (2000). Heuristics for cardinality constrained portfolio optimization. Computers & Operations Research, 27 (13), 1271-1302.
Chen CH. Lu CH. Lin CH. (2020). “An intelligence approach for group stock portfolio optimization with a trading mechanism”. Knowledge and Information Systems, 62: 287-.613
Cui, Tianxiang & Ding, Shusheng & Jin, Huan & Zhang, Yongmin. (2022). Portfolio constructions in cryptocurrency market: A CVaR-based deep reinforcement learning approach. Economic Modelling. 11(9). 89-116.
Deng, D. Yu, D. (2013). "Deep Learning: Methods and Applications Found". Trends Signal Process,7(4): 197–387.
Essid, H. Ganouati, J. Vigeant, S. (2018). "A mean-maverick game crossefficiency approach to portfolio selection: an application to Paris stock exchange". Expert system with applications,1-45.
Fernandez, A. & Gomez, S. (2007). Portfolio Selection Using Neural Networks. Computer & Operation Research, 34(4), 1177-1191.
Ghodusi, S., Tehrani, R. & Bashiri, M. (2015). Portfolio optimization with simulated annealing algorithm. Journal of Financial Research, 17(9), 141-158.
Gulpinar, N., An, L.T.H., Moeini, M. (2010). Robust investment strategies with discrete asset choice constraints using DC programming. Optimization, 59(1), 45-62.
Ma, Yong & Jiang, Hao & Xiao, Weilin. (2021). Tax evasion, audits with memory, and portfolio choice. International Review of Economics & Finance. 71. 896-909.
Mehlawat, M. Kumar, A. Yadav, S. Chen, W. (2018). "Data envelopment analysis based fuzzy multi-objective portfolio selection model involving higher moments". Information Sciences, 460–461:128–150.
Piao, Jinze. (2023). Portfolio Optimization Based on Deep Learning and Factor Constraints. Advances in Economics, Management and Political Sciences. 48(5). 264-273.
Sen, Jaydip & Dutta, Abhishek & Mehtab, Sidra. (2021). Portfolio Optimization Using Deep Learning Models -A Comparative Study of Risk-Based Portfolio Design Approaches. 3rd International Workshop on Machine Learning for Finance, Ca’ Foscari University of Venice, Italy, December 17, 2021At: Venice, ItalyAffiliation: Università Ca' Foscari Venezia 10.
Sharma, Meeta & Shekhawat, Hardayal. (2022). Portfolio optimization and return prediction by integrating modified deep belief network and recurrent neural network. Knowledge-Based Systems. 4(25):128-139.
Shaw, D.X., Liu, S. & Kopman, L. (2008). Lagrangian relaxation procedure for cardinality-constrained portfolio optimization. Optimization Methods & Software, 23(3), 411-420.
Woodside-Oriakhi, M., Lucas, C., & Beasley, J. E. (2011). Heuristic algo rithms for the cardinality constrained efficient frontier. European Journal of Operation al Research, 213, 538e550.
Wu, Songwei. (2023). Dynamic Optimization of Investment Portfolio Based on Self-Attention Mechanism - LSTM Model Prediction. Advances in Economics, Management and Political Sciences. 61(12). 110-119
Zhang, Erchuan. (2023). Portfolio Optimization Strategy Based on Four Deep Learning Models. Advances in Economics, Management and Political Sciences. 47(3). 295-302.
Zhou, X. Wang, J. Yang, X. Lev, V. Tu, Y. Wang, SH. (2018). "Portfolio selection under different attitudes in fuzzy environment". Information Sciences, :264 278-.2.