Machine Learning at Emirates Airlines
Emirates Airlines brings innovative machine learning and data science techniques to the airline industry.
Machine Learning at Emirates Airlines
Why do you think the megatrend you selected is important to your organization’s
management of process improvement and/or product development?
In today’s competitive market, successful airlines have been able to differentiate themselves by offering excellent customer experiences in a cost-efficient manner. Emirates Airlines is at the forefront of customer service and cost efficiency among its competitors (1) (2). Innovation and a forward-looking attitude have been fundamental to achieving this, and machine learning can facilitate its next steps towards enhancing customer experience, all the while reducing the costs of doing so.
Machine learning is suited for predictive tasks such as detecting trends in massive data sets that are correlated to specific effects or events – something that humans would find almost impossible to do otherwise. A problem all airlines face is that of predicting unconstrained demand (3) – this is because as seats fill up, airlines increase the fare and hence constrain demand. Predicting unconstrained demand is important for planning and optimizing fleet allocation to different routes to maximize revenue per unit of available capacity. Similarly, price optimization can be improved through predicting customers’ willingness to pay and then implementing efficient dynamic pricing models (4). Machine learning can also be used predict the effects of weather events (5) or perhaps even airport congestion on flight delays. Maintenance records could be analyzed to detect early warning signs that eventually result in unexpected maintenance delays i.e. condition-based vs time-based maintenance (6). Customer preferences for on-board entertainment could be predicted and viewing suggestions be made to improve the customer experience (7). Other applications include improving customer service response time through automation and predicting passenger baggage size and weight based on each passenger’s history.
All the above applications will result in a differentiating competitive advantage, through enhanced customer satisfaction, lower operating costs, improved revenue management and higher capacity utilization.
What is the organization’s management doing to address this issue in the short term (the
next two years) and the medium term (two to ten years out)?
The management at Emirates have realized that machine learning will be key to unlocking future efficiencies and improving its customers experience. Accordingly, the Oxford-Emirates Data Science Lab was launched, in partnership with Oxford University, to ensure that Emirates remains at the forefront of innovation in this field (8).
It appears that Emirates noticed that the field of data science had not solved the issues it faces to a satisfactory level and so engaged with world experts to solve these issues directly. This demonstrates Emirates proactive approach to remaining at the top of its industry through sustained innovation. This partnership will yield results in the medium term as academic research takes time; however, initial results have already begun to come through in the areas of demand prediction (9) and network modelling to cope with unexpected events and ensure disruptions do not propagate (10). I expect that significant advancements will result from this partnership in the years to come primarily because the reputation of Oxford University is that of academic excellence.
In the short term, Emirates announced plans to implement artificial intelligence initiatives such as: autonomous vehicles to make airport operations more efficient (11), airport baggage handling without the need for human intervention (12) and chatbots for increasing customer engagement in advertisements (13).
What other steps do you recommend the organization’s management take to address this
issue in the short and medium terms?
In the short term, I would recommend that the management increase data collection on all fronts of its business operations. With the pace of advancement in field of machine learning, it’s hard to predict what data will be useful in the future so Emirates should consider collecting new data points across its operations: from maintenance to logistics to customer experience. A larger dataset, across a longer period, will help in the development and training of machine learning algorithms. A larger dataset facilitates the differentiation between signals (real correlations in data) and noise (incidental correlations) through the cross-validation of algorithms against a “test dataset” i.e. data unseen by the algorithm (7). I would also suggest to start testing algorithms early – this would allow fast feedback loops that make the development cycle quicker.
In the medium term, as algorithms are applied to its business operations, I would recommend that Emirates puts suitable checks and balances in place to ensure safety of passengers and employees. For example, from a safety point of view, I would not suggest that condition-based maintenance replace time-based maintenance, but rather supplement it to improve uptime. Continuous efforts should also be implemented to detect and eliminate false correlations that are picked up by algorithms. This can be done by ensuring a team of data scientists maintain an intuitive understanding of the statistical patterns detected by the algorithms are indeed applicable to the context at hand (14) and through the practice of regularization, which reduces the complexity of algorithms thereby making them less likely to follow noise in datasets (7).
In the context of this organization, what are one or two important open questions related to this issue that you are unsure about that merit comments from your classmates?
The research by Oxford University will be published – how can Emirates maintain a competitive advantage if its competitors have open access to this research? Does this mean Emirates considers its real advantage to be executing innovative techniques as opposed to developing them?
(795 words, excluding title, question text & references)
References
- Evaluating Airline efficiency: An application of virtual Frontier Network SBM. Li, Ye, Wang, Yan-zhang and Cui, Qiang. 2015, Transportation Research Part E, pp. Volume 81, pp 1-17.
- Trip Advisor. Emirates Reviews. Trip Advisor. [Online] [Cited: 11 13, 2018.] https://www.tripadvisor.com/Airline_Review-d8729069-Reviews-Emirates.
- Unconstraining Methods in Revenue Management Systems: Research Overview and Prospects. Guo, Peng, Xiao, Baichun and Li, Jun. 2012, Advances in Operations Research, pp. Vol. 2012, Article ID 270910, 23 pages.
- A Machine Leaning Framework for Predicting Purchase by online customers based on Dynamic Pricing. Gupta, Rajan and Pathak, Chaitanya. 2014, Procedia Computer Science, pp. Volume 36, pp 599-605.
- Prediction of weather-induced airline delays based on machine learning. Choi, Sub, et al. Sacramento, CA : IEEE, 2015. 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC). pp. pp 1-6.
- Current status of machine prognostics in condition-based maintenance: a review. Peng, Ying, Dong, Ming and Jian Zuo, Ming. 2010, The International Journal of Advanced Manufacturing Technology, pp. Volume 50, Issue 1-4, pp 297-313.
- Yeomans, Mike. What every manager should know about machine learning. Harvard Business Review. 2015, July.
- Oxford University. Oxford-Emirates Data Science Lab will streamline air travel. [Online] 2015. [Cited: 11 13, 2018.] http://www.ox.ac.uk/news/2015-10-29-oxford-emirates-data-science-lab-will-streamline-air-travel.
- Price, Ilan, Fowkes, Jaroslav and Hopman, Daniel. Cornell University Library. [Online] 11 29, 2017. [Cited: 11 13, 2018.] https://arxiv.org/pdf/1711.10910.pdf.
- Beguerisse, Mariano and Tolvanen, Antti. Dissertation: Spatiotemporal Analysis of Air-Travel Networks. [Online] 12 2016. [Cited: 11 13, 2018.] https://www.maths.ox.ac.uk/system/files/attachments/projects_2016-17.pdf.
- Khaleej Times. Emirates takes off to new world of artificial intelligence. [Online] 2017. [Cited: 11 13, 2018.] https://www.khaleejtimes.com/business/aviation/emirates-takes-off-to-new-world-of-artificial-intelligence-.
- Logistics Middle East. Emirates wants artificial intelligence to handle airport baggage. [Online] 2018. [Cited: 11 13, 2018.] https://www.logisticsmiddleeast.com/business/30871-emirates-wants-artificial-intelligence-to-handle-airport-baggage.
- Liffeing, Ilyse. Emirates Vacations brings a chatbot to banner ads. [Online] 2018. [Cited: 11 13, 2018.] https://digiday.com/marketing/emirates-brings-chatbot-banner-ads/.
- Fedyk, Anastassia. How to tell if machine learning can solve your business problem. Harvard Business Review. 11 25, 2016.
How has Emirates prioritized its machine learning initiative against its different research options? For example, what is the ultimate goal that Emirates is optimizing for, is improving customer experience, reducing cost in manufacturing or improving revenue through optimized pricing. This is important because there are non-trivial costs associated with wide-scale data collection and analysis – particular in a field that is still evolving {1}. Also how does Emirates communicate this approach to its consumers? To what extent should Emirates communicate this as a competitive advantage or should it be seen as a necessary business analytics tool? Finally, have you considered the downside risks of machine learning. You mention the possibility of over-indexing on safety results from predictive maintenance but are there any other potential downside the company should consider?
{1} M. Yeomans. What every manager should know about machine learning. Harvard Business Review Digital Articles (July 7, 2015).
Great article with some solid points about the benefits of Machine Learning in Aviation. While one could spend a significant amount of time addressing all the potential benefits, I’ll focus my comment on the potential to improve maintenance. Reducing maintenance down time is a massive component of maximizing revenue an airline receives from asset turnover. Current maintenance inspection and servicing methodology, both in civilian and military application, is generally based on engine operating time, flight time, or calendar days. While these metrics are based on safety and a general assumption of service life, they fail to take into account dynamic stress on components such as positive or negative g, severe turbulence and loading on wing spars, material contraction and expansion due to extreme temperatures, and the effects of corrosion from salt water or sand in drier climates. If Emirates were able to collect all this information, appropriately apply this information to events experienced by the individual airplane, and create a customized maintenance cycle – it could have a significant impact on asset optimization by addressing service concerns only when needed. To your point, the capabilities and data are becoming widely available. What matters are the metrics a company decides to focus on, those it chooses to ignore, and how it goes about implementing those insights.
Thanks a lot for sharing this article – Really like the practical view on ML in aviation and I’ll come straight to your question:
Yes, I do believe Emirates competitive advantage will eventually lie in executing on the research findings that the University of Oxford Lab provides. Being close to the researchers and showing this strong commitment to Artificial Intelligence in Aviation shows the significance of this technology to Emirates. Other airlines may not have the technology on their radar and may not consider it as relevant, therefore, Emirates can use the closeness to the researchers and their findings along with the actual applications in the industry in order to get a first-mover advantage on certain applications.
On top of that, there might be opportunities for Emirates to co-own some of the patents that come out of the research, considering the investments that Emirates provides to the University.
You mention a vast array of ways that Emirates Airlines can use machine learning to differentiate themselves in the airline market. I agree that there are many applicable uses of this technology, but I was most intrigued by how they can utilize this information to anticipate external factors beyond their control.
Weather is a significant influencer in an airline’s ability to deliver and serve their customer base. Currently, forecasters are able to look at historical trends to determine what they could expect for the future, and this information can be utilized by airlines to bolster their flight offerings in an anticipation of cancellations and delays. However, I think a potential issue that could create a reason for pause would be that researchers often can’t say how or why deep learning algorithms arrive at a given result. Given the lack of tangible outputs and trends, airlines might be more cautious to extend their resources to prepare for events as fickle and seemingly unpredictable as weather.
This is a very interesting read – thanks for sharing! Emirates has long been regarded a leader in the airline industry, and so I am not surprised that it is yet again emerging as a leader in applying machine learning to optimize performance, and in turn maintain its competitive edge in the highly competitive industry in which it operates. To address your questions, I do believe that Emirates’ ability to execute is its main competitive advantage – this is both in terms of their operational capabilities, as well as their vast resources and genuine willingness to invest substantially in maintaining their leadership. I do not believe that competition’s access to this information is a detriment to Emirates as a lot of this technology is soon going to be a commodity anyway. The value will be generated based on how effectively and quickly it is implemented to reap the rewards.
Thanks for the article – In connection to the question, considering that this technology is very new, Emirates will be able to get a first-mover advantage even if competitors can access the research data, especially if Emirates’ concern on innovation and providing a high-quality service to their customers continues to be their priority. Eventually, most airlines will implement ML and only those that prioritize and pursue technological innovation and are able to convert this into a valuable asset, will outstand and maintain their competitive advantage.