Aviation Safety Imbalanced Text Classification Using Machine Learning

Le résumé fourni par la source

To deeply mine the information contained in the aviation safety text and solve the time cost and inaccurate classification results caused by manual classification, natural language processing technology integrating imbalanced data and different feature engineering vectors is used to classify the aviation safety text. Firstly, the Latent Dirichlet Allocation (LDA) is used to reduce the dimension of aviation text. Secondly, weighted processing is applied to imbalanced data texts. Combined with feature engineering vectors of different words or other categories, based on Naive Bayes, Random Forest, Linear Classifier, and an optimized eXtreme Gradient Boosting (XGBoost) model, the practical effects of aviation text mining and classification are studied. The results show that the average accuracy of the classified text after dimensionality reduction, various models that integrate imbalanced data and different feature engineering vectors, has generally improved, with the accuracy rate increased by 6.0%–28.1%, recall rate increased by 12.7%–39.0%, and F1 score increased by 11.4%–29.0%. Among them, the models with better operation time and performance of aviation safety text mining are Naive Bayes of TF-IDF vector feature engineering and XGboost of TF-IDF vector feature engineering after integrating imbalanced text data.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.