Prediction is a key element for decision-making. The information produced by forecasting systems is used in decision-making process. Prediction models can be classified in quantitative and qualitative models based on the level of using mathematical and statistical methods. Prediction in a financial field especially in the case of security pricing is vitally important. The issue of security price behavior and financial asset pricing has been always interesting for investors and academics since the formation of financial markets. Through study and review of available papers in finance, four different approaches are recognizable toward stock price behavior: 1-Technical approach 2-Fundamental approach 3-Modern financial theories approach 4-Chaos and non-linear dynamic approach This research is about predictability of stock price behavior by APT model and Artificial Neural Networks (ANNs) in Tehran Stock Exchange. In order to test our hypothesis, daily stock price of Toseaeh Sanayea Behshar Co. has been selected as a sample. Independent variables consist of five macroeconomic variables, namely, Tehran Exchange Price Index (TEPIX), Foreign exchange rate (Rial / $) ,oil price, gold price and CPI as inflation rate We used multi-factor regression for modeling APT model and perceptron architecture with three hidden layers by back propagation algorithm for AANs model. In order to evaluate the performance of two models, Mean Absolute Deviation (MAD), Mean Square Error (MSE), Mean Absolute Percentage Error (MAPE), Root Mean square Error (RMSE) and coefficient of determination and comparative test of mean error of two models have been used. The results confirm our hypotheses and show better performance for ANNs.