جستجوی مقالات مرتبط با کلیدواژه « AR Model » در نشریات گروه « علوم انسانی »

It is clear that business cycles are inevitable in economy. On the other hand, the economists are always looking for how to form business cycles and so under the effect of economic policies, since the economic situation is depended to these policies. Therefore, the access to more precise business cycles forecasting methods would direct and manage the economic situation and policies powerfully. Hence, the main objective of this study is to construct a new model based on Markov-Switching Autoregressive (MS-AR) model to forecast the business cycles in Iran. In addition, the model constructed is compared to ARIMA to represent its power. GDP data seasonally covers the period 1989: I – 2009: IV collected from Central Bank of Iran. MS-AR and ARIMA models are applied to forecast the behavior of business cycles. By using MAPE, RMSE and Theil criteria (TIC), the results indicate that MS-AR model will work better than ARIMA to forecast GDP business cycles.

This study aims at recognition of appropriate forecasting models for main indices of Iranian stock exchange market including dividends return index, primary and secondary markets price indices, and total market price index. The applied models for forecasting are regression based models including AR, ARMA, TAR, GARCH, EGARCH and GJR-GARCH as well as Artificial Neural Network. The period of study for dividends return and total market price indices are 1378-5 to 1387-8 and 1376-7 to 1386-9, respectively. In the case ofprimary and secondary markets price indices also period of 1381-6 to 1386-12 was used. A period of 12-month was regarded as forecasting period for dividends return and total market price indices while a 9-month period was the forecasting period for primary and secondary markets price indices. The each series data was applied in three form of without adjustment, and adjusted for seasonal and monthly effects. In the case of all indices the artificial neural network showed the highest forecasting error. Based on the findings the forecasting error of the applied regrassion models for dividends return index,primary and secondary markets price indices, and total market price index was obtained %0.72, %2.49, %4.41 and % 5.55 on average, respectively. In general, for dividends return index and secondary markets price index, ARMA, especially after adjusting for monthly effect, was found as the model with the highest accuracy in forecasting. However, in the case of the primary and total market price indices it was cleared that ARCH effect may be useful for forecasting. However, ARMA model forecasts also are of the highly accurate models for primary market price index. But in the case of the total market price index EGARCH was emerged as a model that forecasts more accurately than the other models. On the whole, it was found that adjustment for monthly effectsmay improve the forecasting ability of the most of the selected models.