Predictive maintenance plays a pivotal role in industrial settings, relying extensively on machine learning for the timely detection of equipment failures. However, persistent challenges arise from imbalances and anomalies present in the datasets. Current solutions tend to concentrate on singular aspects, anomaly removal or class imbalance correction without addressing the intricate interplay between the two. This paper introduces the Anomaly Resilient Balancer algorithm, a novel and specialized approach designed to create a balanced dataset free from anomalies. The algorithm leverages robust deviation-aware metrics, employing median absolute deviation and median values to distinguish anomalies from normal instances effectively. This paper not only surpasses existing methods in the context of predictive maintenance datasets but also introduces a groundbreaking, integrated solution for anomaly-resilient balancing. The Anomaly Resilient Balancer sets a new standard for robust machine learning models in real-world applications, showcasing its effectiveness in addressing the complexities of industrial predictive maintenance through a comprehensive and integrated approach. This advancement marks a significant step in developing resilient and reliable machine-learning models for critical applications.
Article Details
Unique Paper ID: 162243
Publication Volume & Issue: Volume 10, Issue 8
Page(s): 470 - 473
Article Preview & Download
Share This Article
Join our RMS
Conference Alert
NCSEM 2024
National Conference on Sustainable Engineering and Management - 2024