As the stock market data is non-stationary and volatile the investors feel insecure during investing. In the recent years lots of attention has been devoted to the analysis and prediction of future values and trends of the financial market. In recent years, mathematical methodology has been used by financial experts and brokers. This study presented Neural Network (NN) approach to develop an efficient model for stock price prediction. Financial ratios were included Earnings Per Share (EPS), Prediction Earnings Per Share (PEPS), Dividend Per Share (DPS), price-earnings ratio (P/E) and earnings-price ratio (E/P) which were extracted from Tehran stock exchange during a decennial period (2000-2009). The training and testing sets were used to develop the NN model. The developed models were subjected to a sensitivity analysis test to assess the relative importance of input variable on model output. Quantitative examination of the goodness of fit for the predictive models was made using R2 and error measurement indices commonly used to evaluate forecasting models. Statistical performance of the developed NN model revealed close agreement between observed and predicted values of stock price, indicates that MLP type NN appears as a promising method for modeling the relationship between financial indices and stock price. The sensitivity analysis indicated that the stock price was more sensitive to DPS followed by EPD, PEPS, E/P and P/E, respectively. In conclusion, the developed NN model could satisfactorily predicted the stock price based on financial indices. Moreover, these models can serve as useful option in determining the relative importance of input variables on model output.