The use of predictive analytics for hydrocarbon exploration in the Denver-Julesburg Basin

We present an approach to predict spatial distribution of a variable from a set of geophysical and interpreted grids using alternating conditional expectations (ACE). This technique is based on nonparametric transformations of the predictor and response variables in order to maximize the linear correlation of the transformed predictors with the transformed response. ACE provides a powerful method to detect underlying relationships between the variables and use them in a regression framework to predict the response variable. A case study is presented that illustrates the approach using a set of grids derived from geophysical attributes (gravity, magnetic, electromagnetic) and interpreted grids (isopach, total organic carbon) as predictor variables to estimate early hydrocarbon production.