ENHANCING RAILWAY STATION SAFETY THROUGH UNSUPERVISED MACHINE LEARNING

Abstract

Railroad operations must be reliable, accessible, maintained, and safe (RAMS) for both passenger and freight transit. Railway station safety and risk incidents are a major safety concern for day-to-day operations in many metropolitan settings. Additionally, the incidents cause harm to the market's brand in addition to expenses and injuries to individuals. Higher demand is putting pressure on these stations, using up infrastructure and raising safety administration concerns. It is recommended to employ unsupervised topic modelling to better understand the factors that contribute to these extreme incidents in order to analyse them and use technology, such as artificial intelligence techniques, to improve safety. Latent Dirichlet Allocation (LDA) for fatality accidents at railway stations is optimised using textual data collected by RSSB, which includes 1000 incidents in UK railway stations. This study offers advanced analysis and explains how to improve safety and risk management in the stations by applying the machine learning topic technique for systematic spot accident characteristics. Through information mining, lessons learnt, and a thorough understanding of the danger posed by evaluating deaths in accidents on a broad and long-lasting scale, the study assesses the effectiveness of text. Predictive accuracy for important accident data, such the underlying reasons and the hot spots at train stations, is provided by this intelligent text analysis. Additionally, the advancement of big data analytics leads to a better understanding of the nature of accidents than would be feasible with a large safety history or with a restricted domain examination of accident reports. High precision and a new, advantageous era of AI applications in railway sector safety and other safety-related domains are provided by this technology.