关键词: 多源文本挖掘, 飞行事故, 关键风险, 负面情绪, 低空空域

Abstract:

In order to enhance the risk control efficiency of civil aviation flight safety in China and accurately identify the key risk factors that cause flight accidents, multi-source text mining of flight accident risks was conducted using Weibo information, news reports, and aviation accident investigation reports as samples. Flight accident risk factors were identified using the bidirectional encoder representations from transformers for topic modeling(BERTopic). The semantic correlation of risk factors was analyzed using the Word2Vec model and complex network, from which the key risk factors were determined. The bidirectional encoder representations from transformers(BERT) was adopted to mine the personnel risk factors that trigger the most serious negative emotions among the public. The key personnel risk factors in low-altitude airspace were identified through word frequency statistics. The results indicate that personnel risk is the key risk factor leading to flight accidents. Among them, failure to strictly follow operating procedures, bird strikes, and sickness of flight crew are the key risk factors affecting flight safety. The non-strict implementation of operation procedures by flight crew not only arouses the most negative public emotions but also constitutes the key personnel risk factor leading to low-altitude airspace accidents.

Key words: multi-source text mining, flight accidents, key risk, negative emotions, low altitude airspace