Predictive modeling is the process of using statistical techniques to predict future events based on various patterns of past and present data. The ordinary least squares (OLS) method is usually used to obtain estimates of model parameters provided that the underlying assumptions of the model are met. The most important of these assumptions that must be verified is the absence of a correlation between the explanatory variables, i.e. a multicollinearity problem, because it leads to ineffective estimators and inaccurate confidence limits, meaning that the model is not suitable for prediction. In this paper, we review several methods for more accurately estimating the parameters of the multiple linear regression model. These methods are: Principal Component Regression Analysis (PCR), Ridge Regression Method (RRM), and Subgroup Selection Method (SSM), which includes several techniques for selecting and excluding variables (all possible subgroup selections, stepwise selection, forward selection, and backward selection). The study concluded that the subgroup selection method according to the forward selection technique achieved the best predictive model [lower root mean square errors (RMSE), mean square errors (MSE) and higher R2] compared to other methods. Based on the best predictive model, the amounts of carbon dioxide emissions were predicted for the period (2022–2030), and the results indicated an increase in emissions rates in Iraq, which requires taking the necessary measures and procedures to reduce this increase.