/multilabel-ULMFiT

Multi-label aviation safety narratives classification

Primary LanguageJupyter Notebook

multilabel-ULMFiT

Multi-label aviation safety narratives classification

Input data: Data for the training and validation datasets were taken from NASA Aviation Safety Reporting System ASRS Database Online. The dataset included passenger flights under FAR Part 121, all records for years 2011 and 2012 for the training and for year 2009 for the validation subsets.

Labels: the contributing factors field

Algorithm: ULMFiT

Tested with fastai v1.028

The notebook presents part of the calculations for a paper "Automated aviation occurrences categorization" (https://ieeexplore.ieee.org/document/8870055) develped initially with an earlier version of fastai)

UPDATE for fastai v2