Data Mining in Aviation Safety Data Analysis

Data Mining in Aviation Safety Data Analysis

Reima Suomi (Turku School of Economics, Finland) and Olli Sjöblom (Turku School of Economics, Finland)
DOI: 10.4018/978-1-60566-230-5.ch011
OnDemand PDF Download:
No Current Special Offers


This chapter introduces aviation safety data analysis as an important application area for data mining. In the beginning of the chapter, the reader is introduced to the basic concepts of data mining. After that, the field of aviation safety management is discussed, and in that connection data mining is identified as a key technology to study through flight incidents reports. Afterwards the test runs for four data mining products, for possible use in the Finnish civil aviation authority, are described in detail. However, before the testing of tools the preparation of the test data for the tools is described in detail. The chapter ends with conclusions that tell that even sophisticated data mining tools are just tools: they do not provide any automatic tools, but skilled users can use them for searching clues in the data.
Chapter Preview

Aviation Safety Data Management

It is widely recognised that air transport is among the safest modes of transport. The global rate of accidents is stabilising. If nothing is done to improve it, the growth in air traffic will lead to an increase in the absolute number of fatal accidents per year. This kind of development would be, clearly, totally unacceptable and therefore new ways of improving air safety need to be explored (European Commission, 2000). Fundamental to every Safety Management System is the principle of collecting and analysing operational data (GAIN Working Group B, 2004). Aviation industry has placed significant investments into collecting aviation safety information from multiple sources. As a result of this process large databases exist, and at the same time, enormous challenges have appeared in analysing the information (Megaputer Intelligence, 2004c).

To improve air safety, the significant allocation of the newest research resources and data processing techniques are unavoidable. Data has been collected, more or less, in different forms during the whole history of aviation. The data collected can roughly be divided into two types: structured data and unstructured, narrative data. The exactness of the incident and accident reports using structured data depends on the number of details and alternatives of the system classification. So, with structured data the explanation of the case usually tells the truth till a certain rate, but completed with narrative data it can reach the level of 100 per cent, at least theoretically. During the execution of various operational and support processes among operations of aircraft fleets large volumes of data are collected. Analysis and review of data is typically complicated requiring human involvement significantly. Usually the data accumulates faster than it can be processed (Wang, Huang, Cao, Shi, & Shu, 2007). The need for automated means to process the data is also increasing rapidly, because the amount of generated and stored unstructured data is increasing fast (Delen & Crossland, 2008).

Complete Chapter List

Search this Book: