National and Kapodistrian University of Athens
Intelligent online optimization algorithms for portfolio analysis and management
Abstract
dc:descriptionOptimization models play a significant role in financial decisions. Popular fields of them are portfolio insurance, portfolio selection, asset allocation, risk management, option pricing, model calibration, etc. and can be solved efficiently using modern optimization techniques. In the literature, due to the high complexity of the mentioned optimization problems, the time-varying (TV) attempts to solve financial problems restrict in solving the corresponding static problems sequentially. Our approach is to define and solve popular financial models in real-time through intelligent online optimization algorithms. The financial models investigated in this dissertation are portfolio selection problems, and they can be classified into two categories. The one category includes TV linear programming (LP) problems and TV quadratic programming (QP) problems, while the other category includes TV nonlinear programming (NLP) problems and TV integer linear programming (ILP) problems.More precisely, the TV minimum-cost portfolio insurance (MCPI) problem is defined and studied as a TV LP problem, while the TV mean-variance portfolio selection (MVPS) problem and the TV Black-Litterman portfolio optimization (BLPO) problem are defined and studied as TV QP problems.In addition, by adding nonlinear constraints, the TV versions of the MCPI and MVPS problems, the multi-period version of the MCPI, and a TV tangency portfolio (TP) problem are defined and studied as a TV NLP problems, while the TV version of MVPS problem is also defined and studied as a TV ILP problem. Note that the nonlinear constraints refer to transaction costs and cardinality constraints.Intelligent online optimization algorithms, which include modern neural network (NN) techniques and state-of-the-art metaheuristics, are employed to solve the aforementioned TV financial optimization problems using real-world datasets.More particularly, the TV LP/QP financial optimization problems are approached by two continuous-time NN solvers. These solvers are the zeroing NN (ZNN) and the linear-variational-inequality primal-dual NN (LVI-PDNN).Moreover, the TV NLP/ILP financial optimization problems are approached by modern metaheuristic optimization algorithms. These metaheuristic algorithms are variations of a popular nature inspired algorithm called Beetle Antennae Search (BAS), whose primary advantage is its low time consumption.The main outcomes of numerical applications in real-world datasets and in various portfolio configurations show that our methods are excellent substitutes for other popular methods. Consequently, our approach will provide a new perspective on TV financial models and will demonstrate an online model solution.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Mourtas, Spyridon
- Μουρτάς, Σπυρίδων
Subjects
dc:subject × 24- Διαχείριση χαρτοφυλακίου
- Μεταευρετικοί αλγόριθμοι
- Νευρονικά δίκτυα
- Μαθηματικός προγραμματισμός
- Portfolio analysis
- Metaheuristic algorithms
- Artificial neural networks
- mathematic programming
- Μαθηματικά
- Φυσικές Επιστήμες
- Εφαρμοσμένα μαθηματικά
- Υπολογιστικά μαθηματικά
- Οικονομικά και Επιχειρήσεις
- Κοινωνικές Επιστήμες
- Χρηματοοικονομικά
- Διοικητική Επιστήμη και Επιχειρησιακή Έρευνα
- Mathematics
- Natural Sciences
- Applied Mathematics
- Computational Mathematics
- Economics and Business
- Social Sciences
- Finance
- Management Science and Operations Research
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/53797