Author: Armando Miguel Nieto Ranero
Programme: Doctoral Programme in Network and Information Technologies
Language: English
Supervisors: Dr Angel A. Juan & Dr Montserrat Guillen
Faculty / Institute: Doctoral School UOC
Subjects: Computer Science
Key words: heuristics, finance, operational research, ALM, portfolio management
Area of knowledge: Network and Information Technologies
Summary
Asset and Liability Management (ALM) has captured the attention of academics and financial researchers over the last few decades. On the one hand, we need to try to maximise our wealth by taking advantage of the financial market and, on the other hand, we need to cover our payments (liabilities) over time. The purpose of ALM is to give investors a series of resources or techniques to select the appropriate assets on the financial market that respond to the aforementioned two key factors: cover our liabilities and maximise our wealth. This thesis presents a set of techniques that are capable of tackling realistic financial problems without the usual requirement of considerable computational resources. These techniques are based on heuristics and simulation. Specifically, a biased randomised metaheuristic model is developed that has a direct application in the way insurance companies usually operate. The algorithm makes it possible to efficiently select the smallest number of assets, mainly fixed income, on the balance sheet while guaranteeing the company's obligations. This development allows for the incorporating of the credit quality of the issuer of the assets used. Likewise, a portfolio optimisation model with liabilities is developed and solved with a genetic algorithm. The portfolio optimisation problem differs from the usual one in that it is multi-period, and incorporates liabilities over time. Additionally, the possibility of external financing is included when the entity does not have sufficient cash. These conditions give rise to a complex problem that is efficiently solved by an evolutionary algorithm. In both cases, the algorithms are improved with the incorporation of Monte Carlo simulation. This allows the solutions to be robust when considering realistic market situations. The results are very promising. This research shows that simheuristics is an ideal method for this type of problem.