A copper price prediction model using weighted least squares regression

The copper is known as an economic indicator because one of the simplest methods of global economic markets analysis is to investigate the price of this developer metal. The extended application of copper in the fundamental parts of any industry and Its developer role in economic activities is the reason for reliability to use this metal for analyzing economics. Any planning necessitates planners to analyze the future circumstances and the only solution to get this target is the prediction. All of the large enterprises like mining companies need price prediction for their decision making about production planning, economic analysis of projects and the new investments opportunities investigations for development. In this article, the copper price is predicted using a weighted least squares multiple regression model. Considering the important role of effective variables recognition in regression models, seven variables including Gold price, Oil and Gasoline price, CIP (Consumer Index Price), GDP, Aluminum price, Euro to USD were used to build the model based on the data from 1988 to 2017. The VIF 1 was calculated for each of the variables to check the multicollinearity, causing to remove Oil price, CPI, and GDP from the model because their VIF value was more than five. In the next pace, autocorrelation was checked and solved using statistical solutions. Finally, heteroscedasticity was checked and the weighted least squares multiple regression model was used to build the model to remove this problem. The adjusted R2 value for this model which is 0.95 indicates the high level of reliability and the precision of the model to predict the copper price.

Keywords:

Copper price , Prediction , Multiple regression model , VIF