developing an appropriate investment model in stock exchange by dea neural network approach

in recent years, the existing competitions between investment companies have been increased largely by entering private investors in capital market. large and powerful companies try to achieve the goals predicted to increase the competition capacity. to analyze the efficiency of investment companies, parametric and non-parametric methods are used. in this research, based on the dissociation power and sensitivity of outliers efficiency frontier in dea, the efficiency of 31 investment companies listed in tehran stock exchange are evaluated by dea models and neural network integrated model as 2 non-parametric methods during 2009-2011. due o the weakness of dea in ranking efficient units, these units will be ranked by anderson and peterson method. in dea neural network integrated approach, multi-layer perspetron network by lm two training algorithm are used. when comparing the results of integrated model and dea, the power of neural network will be represented for evaluating the efficiency.