MetroPT Predictive Maintenance Using Logistic Regression and Random Forest with Isolation Forest Preprocessing

The industry places significant emphasis on predictive maintenance, aiming to anticipate equipment breakdowns and minimize unscheduled downtime. This study utilizes a vast dataset of sensor data to predict equipment failure by employing two well-established machine learning techniques: Random Forest and Logistic Regression. Initial data analysis involves identifying outliers and assessing their potential association with failure before creating the target variable. Model performance evaluation encompasses various criteria, including accuracy, precision, recall, and F1-score. The study’s findings reveal that both the Logistic Regression and Random Forest models exhibit the capability to accurately forecast equipment failure, with the Random Forest model outperforming Logistic Regression in terms of accuracy and F1-score. These findings offer valuable insights for industries seeking to enhance their maintenance procedures and mitigate costly unexpected downtime.