Performance of different clustering methods and classification algorithms for prediction of warning level in aircraft accidents: An empirical study

This paper focuses an overview of the main clustering techniques and classification algorithms for evaluation of risk and safety in civil aviation industry. This paper aim to study the performance of different clustering algorithms is correlated based on the time taken to build model arrangement the evaluated clusters. The Database contains number of accident data records for all categories of aviation between the years of 1950 to 2012. The classification algorithms such as DT, KNN, SVM, NN and NB are used to predict the warning level of the component as the class attribute. The clustering methods are DBSCAN, Farthest First, Filter Cluster, Hierarchical Cluster, Make Density Based Cluster and Simple K Means Cluster and have explored the use of different classification techniques on aviation components data. The rules construct are proved in terms of their accuracy and these results are seen to be very meaningful. This study also proved that the farthest first cluster algorithm will take very few seconds performance better than other clusters on airline data items. This work may be useful for Aviation Company to make better prediction.