Portfolio Selection with Swarm Intelligence

The introduction of budget, cardinality and composition constraints to the portfolio selection problem implies the utilization of modern techniques for the achievement of the solution. In particular, this thesis will analyse Particle Swarm Optimization, a bio-inspired metaheuristic algorithm that aims to explore the search space in order to find optimal solutions. The problem considered consists in the minimization of a coherent risk measure, the expected shortfall, subject to risk adjusted performance constraints, budget, cardinality and fractions constraints. In practice, a chosen number of particles are exploring the set of feasible solutions. To the position of each particle is assigned a value of the objective function which accounts for the risk measure and for penalties associated to the constraints. Particles move according to signals given by their neighbors, by the particle with the best result and by their own memory. The implementation of the PSO algorithm is used to find a feasible and well diversified portfolio composed by Exchange Traded Funds sold on the Italian market, Borsa Italiana.