In recent years, the Big Data and Data Science boom has begun to reach the air transportation industry and we are now encountering the presence of these methodologies in different research projects and programmes, aiming to bring the same benefits to the aviation field that they have provided to other industries. However, the lessons learned from the application of these techniques to other industries need to be taken into account to avoid repeating the same mistakes.
What is exactly Big Data?
Big Data is a term used to describe data sets that are so large and/or complex that traditional data processing applications are inadequate to deal with them. The challenges that arise from Big Data sets include analysis, capture, data curation, search, sharing, storage, transfer, visualization, querying, updating and information privacy. Big Data is well described by these “4 Vs”:
- Volume: sheer volume or scale of the data
- Velocity: frequency of incoming data, particularly for analysis of streaming data
- Variety: different forms of data, both from structured and unstructured data sources
- Veracity: uncertainty of data
Lessons learned from other industries: focusing on using data to obtain value
However, experience in other sectors has led to the conclusion that an additional V needs to be added to these four – and this has led to the use of the new term Smart Data:
- Value: the real objective expected to be accomplished through analysis of a data set
The ultimate objective of any big data project should be to generate some sort of value, to answer a particular business question. Data on its own is meaningless, and in order to harness true benefits, big data needs to be turned into actionable and smart data, with a clear focus on the purpose, (combined) insights, actions and resulting outcomes. Otherwise, it does nothing more than perform some technological task for technology’s sake.
What can this revolution bring to the aviation industry?
For this reason, the aviation industry should start asking about Smart Data projects instead of Big Data analysis. The combination of data scientists and market or sector experts working together is key to solving the particular business question set down as the main objective of any such project. As an example, Smart Data projects could improve the usefulness of collaborative environments among airlines, airports and ATCs by developing adequate predicting algorithms for the key milestones, and balancing the trade-off between algorithm complexity and added value to obtain the optimal solution for each particular case. Along the same lines, ATM predictions both at strategic and tactical levels could be tackled, improving for example Demand and Capacity Balancing methodologies.
An example of Analytics application in aviation: ALG helped EUROCONTROL improve Turn-round predictability
ALG applied these best practices to a particular business question posed by EUROCONTROL: Can we improve Off-block time predictability at European Airports? This question was answered thanks to the joint efforts of airport operations experts and data scientists, who were able to obtain value from EUROCONTROL’s extensive database. This value was in the form of different predictive models which were validated against real flight information. Conclusions extracted from this experience are multiple but it is worth highlighting here that for this particular case, the more complex algorithms were not the most appropriate ones, since even though they were able to provide slightly more added value in terms of average accuracy and error minimization, their computational complexity required too much processing time and historical information to be applied to and cover the whole scope of real-time operations. In conclusion: for this project, the simpler predictive model was the most desirable one.
The example above is one of the different projects in which we have participated applying Smart Data techniques to answer transportation related questions, thanks to ALG’s technology and sector background and experience. The process ALG has followed involves developing different mechanisms to answer one single key question, selecting the best such mechanism through validation exercises that use industry standards (such as the E-OCVM), and balancing algorithmic complexity, computation time and added value to the client.