Giacomo di Tollo

Hybrid metaheuristics for constrained portfolio selection problems

Portfolio selection is a problem arising in finance and economics. While its basic formulations can be efficiently solved using linear or quadratic programming, its more practical and realistic variants, which include various kinds of constraints and objectives, have in many cases to be tackled by heuristics. In this work, we present a hybrid technique that combines a local search metaheuristic, as master solver, with a quadratic programming procedure, as slave solver. Experimental results show that the approach is very promising, as it regularly provides the optimal solution and thus achieves results comparable, or superior, to state-of-the-art solvers, including widespread commercial software tools (CPLEX 11.0.1 and MOSEK 5). The paper reports a detailed analysis of the behavior of the technique in various constraint settings, thus demonstrating how the performance is dependent on the features of the instance.

Hybrid Local Search for Constrained Financial Portfolio Selection Problems

Portfolio selection is a relevant problem arising in finance and economics. While its basic formulations can be efficiently solved through linear or quadratic programming, its more practical and realistic variants, which include various kinds of constraints and objectives, have in many cases to be tackled by approximate algorithms. In this work, we present a hybrid technique that combines a local search, as mastersolver, with a quadratic programming procedure, as slavesolver. Experimental results show that the approach is very promising and achieves results comparable with, or superior to, the state of the art solvers.

Constructing Long/Short Portfolios with the Omega Ratio

We construct portfolios with an alternative selection criterion, the Omega function, which can be expressed as the ratio of two partial moments of the returns distribution. Finding Omega-optimal portfolios, in particular under realistic constraints like cardinality restrictions, requires to solve non-convex optimisation problems. Since standard (gradient-based) optimisation methods fail here, we suggest to use a heuristic technique (Threshold Accepting). The main purpose of the paper is to investigate the empirical performance of the selected portfolios, especially the effects of allowing short positions. Many studies on portfolio optimisation assume that short sales are not allowed. This is despite the fact that theoretically, short positions can improve the risk-return characteristics of a portfolio, and practically, institutional investors can and do sell stocks short.We investigate whether removing the non-negativity constraint really improves out-of-sample portfolio performance under realistic assumptions, that is when optimal weights need to be estimated from the data, different transaction costs apply to long and short positions or short selling is restricted to specific assets.

Hybrid metaheuristic for Portfolio Selection: Comparison with an exact solver and search space analysis

In this paper we use a metaheuristic approach to solve the Portfolio Selection problem, in a constrained formulation which is NP-hard and difficult to be solved by standard optimization methods. We are comparing the algorithms performances with an exact solver and we are showing that different mathematical formulations lead to different algorithms behaviour. Results show that our approach can be efficiently used to solve the problem at hand, and that a sound basin of attraction analysis may help developers and practitioners to design the experimental analysis.

Clustering Local Tourism Systems by Threshold Acceptance

Despite the importance of tourism as a leading industry in the development of a country’s economy, there is a lack of criteria and methodologies for the detection, promotion and governance of local tourism systems. We propose a quantitative approach for the detection of local tourism systems that are optimal with respect to geographical, economic, and demographical criteria. To this end, we formulate the issue as an optimization problem, and we solve it by means of Threshold Acceptance, a meta-heuristic algorithm which does not require us to predefine the number of clusters and also does not require all geographic areas to belong to a cluster.

Distance Measures for Portfolio Selection

The classical Markowitz approach to the portfolio selection problem (PSP) consists of selecting the portfolio that minimises the return variance for a given level of expected return. By solving the problem for different values of this expected return we obtain the Pareto efficient frontier, which is composed of non-dominated portfolios. The final user has to discriminate amongst these points by resorting to an external criterion in order to decide which portfolio to invest in. We propose to define an external portfolio that corresponds to a desired criterion, and to assess its distance from the Markowitz frontier in market allowing for short-sellings or not. We show that this distance is able to give us useful information about out-of-sample performances. The pursued objective is to provide an operational method for discriminating amongst non-dominated portfolios considering the investors’ preferences.

A PSO-Based Framework for Nonsmooth Portfolio Selection Problems

We propose a Particle Swarm Optimization (PSO) based scheme for the solution of a mixed-integer nonsmooth portfolio selection problem. To this end, we first reformulate the portfolio selection problem as an unconstrained optimization problem by adopting an exact penalty method. Then, we use PSO to manage both the optimization of the objective function and the minimization of all the constraints violations. In this context we introduce and test a novel approach that adaptively updates the penalty parameters. Also, we introduce a technique for the refinement of the solutions provided by the PSO to cope with the mixed-integer framework.

A Generalised Linear Model Approach to Predict the Result of Research Evaluation

Peer review is still used as the main tool for research evaluation, but its costly and time-consuming nature triggers a debate about the necessity to use, alternatively or jointly with it, bibliometric indicators. In this contribution we introduce an approach based on generalised linear models that jointly uses former peer-review and bibliometric indicators to predict the outcome of UK’s Research Excellence Framework (REF) 2014. We use the outcomes of the Research Assessment Exercise (RAE) 2008 as peer-review indicators and the departmental h-indices for the period 2008–2014 as bibliometric indicators. The results show that a joint use of bibliometric and peer-review indicators can be an effective tool to predict the research evaluation made by REF.

Local Search Algorithms for Portfolio Selection: Search Space and Correlation Analysis

Modern Portfolio Theory dates back from the fifties, and quantitative approaches to solve optimization problems stemming from this field have been proposed ever since. We propose a metaheuristic approach for the Portfolio Selection Problem that combines local search and Quadratic Programming, and we compare our approach with an exact solver. Search space and correlation analysis are performed to analyse the algorithm’s performance, showing that metaheuristics can be efficiently used to determine optimal portfolio allocation.

Detection of local tourism systems by threshold accepting

Despite the importance of tourism as a leading industry in the development of a country’s economy, there is a lack of criteria and methodologies for the detection, promotion, and governance of local tourism systems. We propose a quantitative approach for the detection of local tourism systems the size of which is optimal with respect to geographical, economic, and demographical criteria: we formulate the problem as an optimisation problem and we solve it by a metaheuristic approach; then we compare the obtained results with standard clustering approaches and with an exact optimisation solver. Results show that our approach requires low computational times to provide results that are better than other clustering techniques and than the current approach used by local authorities.