Background: Linear regression analyses fall into six different kinds namely; simple linear regression, multiple linear regression, logistic regression, ordinal regression, multinomial regression and discriminate analysis(Ghani and Ahmad.2010).Conducting linear regression analysis aims for analyzing and modeling relationships between a dependent variable and one or more independent variables using various techniques. The current study used a stepwise multiple regression which is known as a combination of forward selection and backward elimination method. Objective: The study reported in this paper mainly aimed at selecting the suitable controlled variables in the forecast Malaysia’s imports. The study will be limited to six variables which are the exchange rate, producer price index of imports of (MT), G.D.P, the value of exports of (MT), the average of tariff tax of imports of (MT), the average sales tax of imports of (MT). According to the data available, the time frame for this study will be determined by using quarterly data covering the period from 1991 to 2013 (the period of building the model). Results: Based on the results obtained from the stepwise regression method, It was found that the dependent variable follows the normal distribution with the level of significance 0.01.The four multiple regression models were also estimated and were found to be all good in terms of the coefficient of determination R2.It was found that the first and fourth models are good in terms of VIF, but the first model is the best among all models. By performing the test of Durban Watson, results showed that the linear model suffers from the problem of autocorrelation in residuals. However, after treating the model and solving the problem, we obtained an effective model which does not suffer from this problem and is capable of prediction, and consistent with the economic theory in terms of signal parameters. It was also found that the first model is good in terms of its predictive ability. Diagnostic measures showed that the model is very suitable for predicting. Conclusion: it was found that only three controlled variables which are G.D.P, the value of exports and the average of Tariff tax were selected in this study, and consistent with the economic theory in terms of signal parameters. This indicates that only these variables affect the value of imports of (MTE) in Malaysia.

Abstract: Objective: With the increasing importance of imports as one of the important factors of economic growth, the current study proposed techniques of more reliable and predictable Malaysian imports of crude material in the future. Specifically, this study proposes composite models for probabilistic imports of crude material forecasting in Malaysia. Methodology: In this study, the proposed composite models (With regression processing of heteroscedasticity), (With regression processing of heteroscedasticity and autocorrelation) were employed to extract information that assists in increasing accurate forecasting of the size of the Malaysian imports as well as forecasting engines and compare it with other commonly used models including regression models and ARIMA models. Results: The forecasting results of the study showed that the composite model (With regression processing of heteroscedasticity) approach provides more probabilistic information for improving forecasting of Malaysian imports of crude material. Conclusion: The results also showed two sets of benefits: The main benefit is that the composite model (Without regression processing) is capable of solving the problem of autocorrelation in residuals but it was unable to solve heteroscedasticity in the residuals. The second benefit is processing the problem of autocorrelation in the composite model in a case when it is not processed in the regression model. However, in the case of the emerging problem of the heteroscedasticity, it can be processed in the regression model prior to the composite model formation.