Portfolio selection is one of the most important problem in financial issues. It is significant to apply bourse market conditionsin portfolio selection model.in this thesis,we added some important factors like cardinality constraints,transaction costs,and the minimum transaction lot simultaneously to get it more closer to the real conditions of the problem. Uncertainty of the model parameters is another very important factor to achieve optimum of the portfolio selection. There are various approaches to encompass uncertain parameters, so that the most important of them consist of sensitivity analysis,stochastic programming, and robust optimization. In this thesis,we use robust optimization to model the problem. The problem shifts to the MIP(mixed-integer problem) according to the minimum transaction lot and cardinality constraints and it is not possible to solve the problem by exact procedures.Due to this characteristic of the problem,we propose a meta-heuristic procedure to solve the problem approximately.in this paper, we use two different types of uncertainty sets for parameters,cardinality, and norm set (introduced by Bertsimas),to expose the effect of defined set type for uncertain parameters on the portfolio and compare with each other. At last, several models of portfolio selection were compared with respect to various uncertainty sets. The results have shownthat the efficient frontier of robust counterpart with respect to the uncertainty set with cardinality constraint in a similar level of probabilistic guarantee and expected return, have higher risk than the robust counterpart, according to D-Norm set.In addition, robust counterpart with respect to D-Norm uncertainty set in a similar level,has better turnover than probabilistic guarantee.In other words, it shows less changes about risk than other robust models.