The science of geology has a long history. It forms the basis for traditional geological metrics, such as predicting oil production from shale wells from geologic and completion rate parameters. In the past, calculations were driven from first-order geological principles developed using physical models. In some cases, these relationships were developed from established linear regression models.

Nowadays geologists are learning to make use of more advanced techniques originally developed for Artificial Intelligence applications, and now modified for solving geological and petroleum problems. These include cluster analysis, artificial neural networks, decision trees text mining, and many resampling methods such as random forests and ensemble models.

This talk describes the cross-over between some of the AI algorithms and modern geological metrics. The techniques are described in the contest of AI and then illustrated using problems associated with forecasting oil production from geological and well attributes.