Tesi di LAUREA SPECIALISTICA Titolo Ottimizzazione di portafoglio con Conditional Value-at-Risk: modello ed euristica Data 2010-05-03 Autore/i Mazza, Maria Cristina Relatore Amaldi, E. Relatore Nencini, M. Full text non disponibile Abstract In this thesis we study and solve an asset allocation optimization problem using methodologies peculiar of operational research. The model considered is that proposed by Rockafellar e Uryasev ([26]). The aim is the minimization of -Conditional Value-at-Risk, a coherent risk measure that represents the expected loss exceeding -Value-at-Risk, under constraints of minimum return, a minimum limit and maximum limit to the weight of each asset and an upper bound to the number of assets included in the portfolio: this latter constraint is imposed in order to reduce, for example, transaction costs and it makes this problem not convex any more. We analyse two cases: in the rst one we consider a small number of assets and we imagine that the portfolio is subject to market risk, while the second portfolio is bigger and it is exposed to credit risk. The model is solved with GLPK, a solver for linear mixed integer problems; because of long solving times we propose and develop an effcient Tabu Search heuristic. Also we discuss the results comparing the quality of optimal solution and execution times: the heuristic gets to solutions that diverge from GLPK solutions with an error smaller than 1% and the times are comparable or lower than the ones of GLPK.