Different Machine Learning Methods for Predicting SAGD Production Performance

Ziteng Huang
截屏2020-12-07 下午2.55.49

Abstract

The different machine learning methods and different data samples are tested in this study. A larger data sample number and greater randomnessof a data sample has better prediction accuracy in the machine learning training.  In this study, the XGBoostand LightGBM models have better prediction performance than the ANN model. But if the data sample number or the data randomness is not great enough, the ANN model can still have ability to predict new data while other two models (XGBoostand LightGBM) have worse performance on it.