Multivariate Prediction Based on the Gamma Classifier: A Data Mining Application to Petroleum Engineering

A novel associative model was developed to predict petroleum well performance after remedial treatments. This application is of interest, particularly for non-uniform oilfields such as naturally fractured ones, and can be used in decision support systems for water control or candidate well selection. The model is based on the Gamma classifier, a supervised pattern recognition model for mining patterns in data sets. The model works with multivariate inputs and outputs under the lack of available data and low-quality information sources. An experimental dataset was built based on historical data of a Mexican naturally fractured oilfield. As experimental results show, this classifier-based predictor shows competitive performance compared against other methods.