Machine Learning and Knowledge Extraction to Support Work Safety for Smart Forest Operations

Forestry work is one of the most difficult and dangerous professions in all production areas worldwide - therefore, any kind of occupational safety and any contribution to increasing occupational safety plays a major role, in line with addressing sustainability goal SDG 3 (good health and well-being). Detailed records of occupational accidents and the analysis of these data play an important role in understanding the interacting factors that lead to occupational accidents and, if possible, adjusting them for the future. However, the application of machine learning and knowledge extraction in this domain is still in its infancy, so this contribution is also intended to serve as a starting point and test bed for the future application of artificial intelligence in occupational safety and health, particularly in forestry. In this context, this study evaluates the accident data of Osterreichische Bundesforste AG (OBf), Austria's largest forestry company, for the years 2005-2021. Overall, there are 2481 registered accidents, 9 of which were fatal. For the task of forecasting the absence hours due to an accident as well as the classification of fatal or non-fatal cases, decision trees, random forests and fully-connected neuronal networks were used.

Automated summary

Forestry work is one of the most difficult and dangerous professions, and occupational safety is crucial for sustainable development. This study evaluates accident data of Austria's largest forestry company and explores the use of machine learning and artificial intelligence in predicting absence hours and classifying accidents as fatal or non-fatal.