Dataset for drone problem identification and severity estimation

The paper proposes DroSev, a dataset for drone problem identification and severity estimation. The collection of drone flight log messages was acquired from publicly accessible sources on Mendeley Data and AirData. This dataset consists of two subtasks: binary problem identification and multiclass problem severity classification. The former task used only the collection of log messages from Mendeley Data, and the latter task used the merged collection of log messages from both sources. Each subtask has a train and test split with an 80:20 ratio generated with stratified sampling. Further syntactical characteristics are reported and summarized.