Enhanced prediction of hard rock pillars stability using fuzzy rough feature selection followed by random forest

Abstract

Pillar stability in underground hard rock mining task is one of the most challenging safety problems to be determined during mining task. This stability analysis requires proper input variables, which are also known as parameters. The prediction of pillar stability is a key task for which various machine learning based methodologies are available in the literature. In this study, we present a novel methodology to enhance the prediction of the stability of hard rock pillars by using fuzzy rough feature selection with rank search and evolutionary search. Initially, irrelevant and redundant features are removed, using fuzzy rough feature selection technique. Thereafter, machine learning techniques are used for reduced dataset and the findings are recorded. Then, fuzzy rough attribute evaluator is deployed to present the rank of different features according to their influence. The work presents schematic representation of the proposed methodology. Finally, a comparative study of the proposed approach with the existing techniques is presented. From the work and discussion, it can be observed that random forest (RF) is producing the best results till date as the average accuracy produced by present approach and existing approach are 83.3% and 79.2% respectively with percentage split of 80:20. © 2019, Books and Journals Private Ltd. All rights reserved.