Database

In 2024, the forecasted physical availability of heavy equipment at PT. BMC was 89%, using a simple average method, while the actual values ranged from 85.33% to 89.92%. The average difference between the forecast and actual values was -2.02%, with the largest gap occurring in the month with the lowest actual value, at -3.67%. This inaccuracy indicates an over-estimation bias that could impact operational planning. If the issue of misforecasting continues, the company will face challenges in predicting or planning unit replacements. Consequently, declines in availability will become unpredictable, leading to consistently low equipment availability and disruptions in operations. This study aims to train a regression-based machine learning model to identify the model with the best evaluation performance based on field parameters such as equipment condition, equipment age, equipment hours meter, maintenance frequency, equipment activity, and historical maintenance records. This approach will provide more accurate predictions aligned with actual field conditions. PT BMC will also be able to identify the parameters closely related to availability using multivariate analysis. Based on this information, recommendations for improvements will be provided to enhance the availability of heavy equipment. The study's findings indicate that the best-performing model for forecasting availability is the decision tree algorithm, with an R² train score of 1.000 and R² test score of 0.8624, RMSE train of 0.000, RMSE test of 0.0627, and MAPE train of 0.0000 and MAPE test of 10460471212418. A critical parameter significantly affecting availability is risk ranking, which shows a negative correlation with availability, with an R² value of -0.8973. Based on these findings, the recommendations for PT BMC include increasing equipment cleaning, focusing attention on units operating in high-risk areas, and implementing predictive and preventive maintenance procedures.