We have researched the applicability of machine learning for identifying work operations and examined which machine learning methods yield the best results. To this end we conducted a data analysis of existing data on a cable yarder using machine learning methods. Several models were developed with precise configuration of components and evaluated their reliability in classifying the machine’s operational status. Special emphasis was placed on identifying productive time and extended downtimes, which enables assessment of the crane’s availability and utilization. The results indicate that the Random Forest algorithm is the most successful in classifying operations and predicting key indicators. It achieved the highest accuracy and the lowest mean squared error, with no significant differences between predicted and actual results. The Decision Tree also demonstrated high reliability, while the Neural Network lagged behind due to low classification accuracy. In determining the extraction dlistance, we observed discrepancies between manually obtained data and machine-reported data, which require further investigation. Machine learning proved to be an effective tool for predicting the yarder’s operational activities.
|
