I develop a two-step subset selection procedure to extract the best-performing predictors from a large dataset and combine them to identify a set of best-performing models. I apply the methodology to build an index to forecast the probability of having the euro area year-on-year inflation below the 2% level in a medium-term horizon--i.e., the Deflationary Pressure Index (DPI). I compare the index with the probabilities reported in the European Central Bank Survey of Professional Forecasters (ECB SPF) and show that, although the indexes are comoving, the DPI is more operationally convenient and timely in catching the inflation turning points. As a final exercise, in a real out-of-sample forecast, the index shows that having medium-term inflation above the 2% level before 2019 is unlikely. .