Evaluating the Performance of Machine Learning Algorithms in Predicting Industrial Equipment Maintenance Costs

Abstract

Effective maintenance planning for agricultural machinery requires a reliable estimation of future repair and operating costs. Tractors are among the most important capital assets in agricultural operations, and their repair, oil, and fuel costs tend to increase as cumulative operating hours grow. Accurate prediction of these cost components can support maintenance budgeting, economic life assessment, and replacement-related decision-making. This study applies a polynomial regression-based machine learning approach to predict indices of agricultural tractor maintenance costs. The dependent variables include the cumulative repair cost index, cumulative oil cost index, and cumulative fuel cost index, while cumulative operating hours are used as the main explanatory variable. Cost values are standardized using cumulative cost indices to reduce the effects of inflation, differences in tractor purchase prices, and variations across tractor types. The predictive performance of second-degree and third-degree polynomial models and an exponential model is evaluated using goodness-of-fit and error measures. The results indicate that the third-degree polynomial regression model yields the lowest prediction error among the cost indices studied. The findings suggest that polynomial regression can provide a simple, interpretable, and practically useful tool for forecasting tractor maintenance cost trajectories and supporting machinery management decisions.