Towards Big Data Analytics and Mining for UK Traffic Accident Analysis, Visualization & Prediction

Road traffic accident (RTA) is a big issue to our society due to it is among the main causes of traffic congestion, human death, health problems, environmental pollution, and economic losses. Facing these fatal and unexpected traffic accidents, understanding what happened and discover factors that relate to them and then make alarms in advance play critical roles for possibly effective traffic management and reduction of accidents. This paper presents our work to establish a novel big data analytics platform for UK traffic accident analysis using machine learning and deep learning techniques. Our system consists of three parts in which we first cluster accident incidents in an interactive Google map to highlight some hotspots and then narratively visualize accident attributes to uncover potentially related factors, finally we explored several state-of-the-art machine learning, deep learning and time series forecasting models to predict the number of road accidents in the future. The experimental results show that our big data processing platform can not only effectively handle large amount of data but also give new insights into what happened and reasonably prediction of what will happen in the future to assist decision making, which will undoubtedly show its great value as a generic platform for other big data analytics fields.