This paper examines whether the inclusion of several commodity indexes in multivariate mod- els could improve the forecast of CPI inflation for Belgium and Norway. The results from the multivariate forecast models are compared to three different univariate models: an AR(1), an Atkeson & Ohanian model, and an other autoregressive model with specific lags chosen. The second objective of this thesis, is to assess if the economic profile of a country would provide better forecasting results. In this case, due to the importance of commodities for the Norwegian economy (almost 70% of their total exports), would the forecast results be better than those of Belgium ? The variable forecasted is the consumer price index inflation while the commodity indexes originate from two sources: the Commodity Research Bureau (CRB) and the International Monetary Fund (IMF). In total, four commodity indexes are chosen: two generic indexes and two sub-indexes. The data set ranges from January 1992 till May 2015. The forecasts are performed at four different time horizons. The root mean square error (RMSE) is the main indicator used to compare the results. In some case (for Belgium), the multivariate models with the commodity indexes provide better results than the univariate benchmark models. However, those results are not significant enough to affirm that the inclusion of commodity indexes provides more accurate forecasting results than univariate models. Finally, the fact that commodities represent an important share of the total export of Norway does not seem to provide better forecasting results. ; Master [120] en Ingénieur de gestion, Université catholique de Louvain, 2015