Machine Predictive Maintenance Classification Using Machine Learning
Dharithri B Sharma, Sripradha, Nikita Nikita, Ashwini Kodipalli, Trupthi Rao, B R Rohini
Abstract
Machine Learning Predictive maintenance (PdM)/condition-based maintenance algorithms are pre-defined models that let you perform pre-processing on any dataset and detects possible equipment failures with suggestions on actions to take to prevent these failures. It is a difficult task to identify high maintenance software modules or classes in the early stages. Many Software Maintainability Prediction (SMP) models have been constructed using machine learning (ML) and ensemble methods. But the main problem with these models is low accuracy of prediction due to the imbalanced datasets used for training the models. Our main focus is to make a comparative study of different classification models applied on predictive maintenance dataset and identify which one gives best performance in terms of its accuracy and other performance measures. We measured the predictive power of the different machine learning algorithms by taking into consideration the accuracy score values. The main goal of this analysis is to improve equipment reliability and reduce cost and downtime associated with unexpected equipment failures. The various machine learning algorithms were applied and observed that Random Forest has outperformed with the accuracy of 98%.