Anomally Detection of Airways Transportation using Deep Neural Networks and Feature Selection Based Particle Swarm Optimization Algorithem

Abstract

Aviation industry is an important part of transportation system in any country. Damage to this industry cussed expensive and sometimes Irrecoverable problems for governments, so for scientist in various field is attractive for research in aviation. There are several unpleasant events in aviation that can be due to a variety of human error, unwanted defect hardware or terroristic and criminal actions that happen eventually with change in the behavior of a normal airplane. Often new method of detecting anomalies in finding these events leading to useful result for this industry. There are many limitations for diagnosing abnormalities based on human supervision. One of these limitations is human error. This problem after long working hours is more prevalent. Greatest limitations of using manpower is that simultaneous checking of all the flights according to the high density transportation in main airports is not possible. Proposed methods for automatically detecting anomalies in this paper is based on using one of the powerful tools in artificial intelligence and computer science called neural networks. The method proposed in this paper used neural networks for create an efficient model for all data. Sure using feature selection based particle swarm optimization helped us to achieve best result. In this paper for evaluate method we use data from three major airports including Tehran's Mehrabad airport, Mashhad’s Hasheminejad airport and yazd’s Shahid Sadoughi airport. Simulation results shows proposed method is reliable and can help us for reduce accidents caused by human errors, which achieving 99.78 percent accuracy rate clearly shows this fact.