Machine Learning: Leveraging FAA METAR Data to Predict Weather via Sci-kit Learn; Identifying When to Compete or Collaborate in Aviation Research and Industry

Khatod, Atharv, School of Engineering and Applied Science, University of Virginia
Neeley, Kathryn, EN-Engineering and Society, University of Virginia
Graham, Daniel, EN-Comp Science Dept, University of Virginia

Advances in technology and navigation have allowed major developments in aviation such as autopilot, auto-landing and throttle systems as well as a slew of safety systems. As more computational power becomes accessible, new research shows avenues to improve airlines’ operational efficiency such as itinerary matching via machine learning, or a new algorithmic design to book airport slots efficiently. As the rate of technological progress increases, so do the advances made in the aviation industry. Whether research that can harness machine learning and predictive algorithms for the benefit of airlines should happen jointly between airlines or be separate in a competitive fashion is what my STS paper explores. My technical topic ties in closely with my STS topic by focusing on a practical application of an aspect of machine learning that could be applied to airline operations: weather prediction. It’s important to distinguish between what types of research benefit in disjoint versus collaborative environments, this requires a sociotechnical approach to a core technical problem. Understanding who this research will impact and in what manner are key factors in determining the environment of development.
The technical portion of my thesis produced a basic METAR (Meteorological Aerodrome Reports) prediction framework centered around Dulles Airport via machine learning.
METARs are used actively by airlines and airports to monitor weather changes at a given airport. Pilots use these reports often in flight to configure their aircraft appropriately for oncoming weather conditions. METARs are provided hourly at all major US airports and
occasionally at smaller ones. The machine learning model we built was able to provide ~90% accurate predictions of temperature, dew point, general wind speed and direction, 24 hours ahead of the current time during the summer months. This basic model and the method of real-time
data collection of METAR data allows for a basic avenue for further utilization of this framework at other airports.
In my STS research, I explore the benefits and drawbacks of whether airlines should collaborate or compete when it comes to researching avenues to improve operational efficiency,
customer experience or safety. This analysis is done via exploring various research papers proposing solutions to operational challenges that airlines face such as congestion and delays at airports, or inflight turbulence. I decided to utilize a collaborative-competitive research framework in examining the sociotechnical and innovative effects of a specific avenue of research - this influenced the decision on whether there is greater utility in collaborating or competing. Through my research, I found that there’s a benefit to both competitive and collaborative approaches but in different aspects of what that research accomplishes. Research aimed at improving safety and security should be collaborative as all airlines can provide their data and perspectives on the issue whereas something like itinerary matching should be a competitive process, with the better airline attracting more customers, pushing each airline to develop a more competitive product.
My STS and technical topics are very closely related. The technical paper was one avenue of
implementation of predictive machine learning to make aviation a more efficient industry via weather prediction. I wouldn't have learned how realistic this project was without the technical side of it. With the STS paper delving into more research aspects of machine learning, having a technical product to showcase made the STS topic feel more tangible. Conversely, when working on a technical project, it’s easy to be satisfied with what you’ve developed and forget to analyze the societal implications of this developed project. With STS focusing on the ethical development of technology, I found that viewing the research and its applications with a sociotechnical perspective allowed me to determine where business competition should be set aside for the common good - especially when regarding safety in aviation or when interacting with government agencies.

BS (Bachelor of Science)
Aviation, Competition, Collaboration, Airline Industry, Aircraft

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Daniel Graham

STS Advisor: Kathryn Neeley

Issued Date: