Professional Documents
Culture Documents
Applying Data Mining To Extract Damage L PDF
Applying Data Mining To Extract Damage L PDF
Research Paper
Applying data mining to extract damage level of bird strike on helicopters
Camelia Arbabzadeh*1, Abbas Toloie Eshlaghy2
1Department of Information Technology Management, Science and Research Branch, Islamic Azad University, Tehran, Iran.
2Department of Industrial Management, Science and Research Branch, Islamic Azad University, Tehran, Iran.
Available online at: www.IJSMD.Com
Received 20th November 2014, Revised 20th December 2014, Accepted 28th December 2014
Abstract
Underestanding the risk factors of damage will be helpful to mitigate damage of bird strike. Planning for preventing bird strike
evidence need to risk factors and relation of these factors to level of damage. Hence, the goal of this paper is extracting the risk factors
by consideration of damage level imposed on helicopters. The data source of the current research is the Federal Aviation
Administration Wildlife Strike Database which was gathered for about twenty years. In this paper, we used four kind of decision tree
algorithms included C5.0, CART, C&R TREE, and QUEST to extract damage level of bird strike (D, S, M, and N that demonstrate of
destroyed, substantial, minor, and not defined) on helicopters. The results demonstrated that bird size, helicopters type, phase of flight,
speed, and type of engine are the most important risk factors causing damage to helicopters, respectively. Finally the research
concludes with some suggestions to reduce damage of bird strike on helicopters.
Keywords: Bird Strike; Data mining; Decision tree.
Introduction information repositories likes data warehouses or databases
A bird strike, sometimes called birdstrike, bird ingestion (if (Gürbüz et al.2011) (Jiawei et al.2001).
bird swallowed by engine of aircraft), bird hit, or BASH (for
Bird Aircraft Strike Hazard) is a collision between an airborne Literature Survey
animal (usually a bird or bat) and a human-made vehicle, Both (2010) for a given period of time a risk assessment
especially aircraft (wikipedia). matrix constructed which is included bird species according to
Bird strike has become a major threat to air safety globally. Over frequency of bird strikes are involved and the percentage of
the years, collision between birds/wildlife and aircrafts has strike resulting in damage. The matrix is appropriate tool for
resulted in the death of hundreds of people (USMAN.2012). The manage and monitoring the bird strike hazards (Both et al.2003).
cost of bird strikes estimate of US$ 1.2 billion per year that this Allan (2006) introduces the well-known probability-severity risk
cost is with concerned of damage and delays (Allan ET assessment matrix for bird strike risk assessment that the bird
AL.2000). ICAO recommended to register the data of airport strikes per bird species over a given 5 year period is considered
bird/wildlife strike and organizing a programme for control and a measure of strike probability for a given airport. Damage is a
preventing this kind of events.This organization also introduce measure of likely severity that is the proportion of strikes with
roles and responsibilities within a bird/Wildlife strike control each 4Pspecies (Allan et al.2006).
programme (Airport Services Manual). Hence, quantify This risk assessment matrix is used as auditing tool for the
important parameters that lead us to assess level of damage that bird strike prevention activities of airfields (Searing.2005). It is
probably will occur in the result of bird strike is necessary. a simple tool which shows for which bird species further action
On January 15, 2009, the plane struck with geese, lost engine is needed. The risk assessment matrix always need to support
power and landed in Hudson river after 3 minutes into the flight. with a reporting system that furnish reliable information about
These bird strike events have major effect on safety in aviation occurred bird strikes (Both et al.2003) (Linnel,1999).
industry. Many researches try to find effective methods and Risk is defined as the product of the severity and probability
make suggestions to put an end to these kind of events but of wildlife strikes during a predefined period (Allan.2001).
applying data mining methods on bird strike is rarely occurred. Wildlife risk in and around airport is a concerning subject in
With the rapid growth in air travel, considering risk and safety is airports, need to find a threat to measure risk of bird strike in
important in aviation (Nazeri.2002). Aviation industry is one of airports. To provide a way posed by each wildlife species,
the fields that data mining methods have been applied (Gürbüz wildlife hazard management experts have suggested to utilize of
et al.2011). wildlife risk hazard assessments. In considering that risk is
Data mining is finding relationships to summarize the data in defined as the result of the severity and probability of wildlife
ways that exploit effective data that is useful to the data owner. strikes during a period of time (Allan.2001). Dolbeer et al
The relationships and summaries derived through a data mining (2000) described that birds with bigger size had more
exercise are often referred to as models or patterns (Gürbüz et probability to provide damages more severely than others in bird
al.2011) (Hand et al.2001). strike accidents. Measure of the severity is using the size of
knowledge discovery in databases (KDD) is another popular birds and the probability is the frequency of bird strike
name of data mining that is automated extraction of patterns and occurrence (Dolbeer at al.2000).
can show knowledge implicitly that existed in massive Tan et al (2010) for measuring an aerodrome’s wildlife
safety performance show a basic framework on the practical
*Correspondent Author : Camelia Arbabzadeh (camelia.arbabzadeh@gmail.com) application of wildlife risk hazard assessments. The case study is
Manuscript No: IJSMD-KINA-2014-506 Changi Airport and in this research try to help airports in better
Fig 1. Diagram of the model provided for prediction of damage level caused by bird strike
International Journal of Scientific Management and Development
918
International Journal of Scientific Management and Development ISSN:2345-3974
Vol.3 (2), 917-922 February (2015)
5.Evaluation: in phase of evaluation must determine if the parameters according to output parameters. In the next step,
quality of constructed model is acceptable or need to some valuable inputs are obtained using the rules provided by derived
modifications. In this part for evaluating the model, used decision trees. As the figure 2 shows, helicopters type, bird size,
evaluation node for training and test parts of data and the result phase of flight, and engine type are the most important
was accommodating. parameters according to C&R Tree, respectively. The
helicopters type and bird size show a much more significant
Results and Discussion effect compared to two other parameters. As seen, helicopters
As it was pointed out in the previous section, feature speed has not had any considerable influence on the damage
selection is used for determing the most important input imposed on helicopters.
Conclusions namely C5.0, CART, C&R TREE, and QUEST have been used.
In this paper, it is attempted to model the bird strike to IBM SPSS modeler, a data mining software which possesses
helicopter incidents using the FAA wildlife database. Since the different modeling techniques, has been used as the modeling
decision trees are suitable prediction tools, four decision trees tool. In order to create the model, the data were first cleaned and