A Model for Testing and Improving Stock Market Efficiency

This study provides a model for testing stock market efficiency. Case study of this research is Tehran Stoch Exchange (TSE). To test the efficiency of TSE, Neural Network capable of learning the underlying dynamics of complex processes are used. After an explanation of the underlying theory, Neural Networks are applied to a trading simulation based on predicted value of 2 indices, over the period of Mehr 1383- Esfand 1385. An array of trading results was derived using the predicted index values and different trading parameters, four threshold levels on which tradind decisions were based, and four transaction cost levels. The simulation results indicate that in most cases, neural Networks were able to make statistically significant excess return over the passive "buy and hold strategy". This suggests that the stock market might be inefficient. For validation of this model, "run test" was used to prove that the results of simulation are correct. At last, this research provides some good ways for improving stock market efficiency.