PROACTIVE SAFETY PARADIGM FOR AVIATION SECURITY USING A HYBRID MODEL OF SVM AND DNN

Abstract

Due to the growth in air traffic, there exists a dire need for inculcating a system which not only alleviates the load on the airline operators to assess the associated risk during the process of flight aviation but also developing a ‘proactive safety’ paradigm which helps in assessing the severity of abnormalities, by categorizing the various parameters as “high, medium or low level” risk factors. To implement the aforementioned, a predictive model needs to be developed to examine a wide variety of possible cases and provide for a crisp value for the entailing consequences associated with the possible outcome. The dataset is taken from the ASRS online database repo, maintained by NASA to record all the aviation mishaps. First, we categorize all the events, based on the level of risk associated with the event consequence, into three groups as mentioned earlier, followed by a support vector machine model to find the relationship between the “event synopsis (text format)” and the “event consequence” by using tokenization techniques. In parallel, a deep neural network is trained to map the coupling between event contingent features and event results. Furthermore, a strategy to agglomerate the results from the two different models is deployed thereby improving the prediction. Finally, the prediction on risk level categorization is extended to event-level outcomes through a probabilistic decision tree together with a performance evaluation metric to display the efficiency of the said proposal.