A Custom Stacking-Based Ensemble Learning Approach to Predict Failure of Stripper Well

Prediction of equipment failure has always been a challenging task. Analytical and statistical approaches for prediction of equipment failure have been employed for a long time. Analytical approach is based on criterion, while statistical approach is data driven. Despite its accuracy, statistical approaches fail with large data entries having high dimensionality. Advanced machine learning techniques come to rescue. In this study, an effort has been made to predict failure of stripper well with classical machine learning algorithms followed by a custom stacking-based ensemble learning approach. Classical machine learning algorithms like Support Vector Machine, K-Nearest Neighbour, Logistic Regression, Gradient boosting etc. have been applied to predict failure instance. Micro F1-score has been selected as a measure of prediction accuracy. A novel custom ensemble machine learning approach has been implemented to obtain better prediction accuracy compared to previously applied algorithms. Proposed novel approach has successfully predicted classification case of failure with micro F1-score of ~0.9887. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd