Client Overview

SITA is the world’s leading specialist in air transport communications and information technology, with nearly every passenger flight relying on SITA technology.

Challenge

SITA are currently exploring the extent to which AI and Machine Learning can help to solve some of the challenges faced by the air travel industry. We are exploring a number of interesting areas with SITA, including:

  • Mishandled baggage – how much can we learn about a bag from a photo?
  • Disruption – can we train a model to aid humans to find the optimal stand for an aircraft to use?
  • Queue times – how long do passengers spend in the security queue and how can we use this data to improve the experience?
  • Aircraft turnaround – can computer vision help us to detect important events during an aircraft’s turnaround, and therefore to understand any likely delays?

Our approach to these sorts of challenges allows organisations to minimise risk before embarking on a time-consuming and costly full-blown AI project. Before committing to any sort of results or deliverables, we need to review the data that a client has, so we first recommend a Data Study.

A Data Study is a relatively small-scale, fixed fee activity where our data science team explore and start to draw insights from our client’s data. The aim is to assess the feasibility of using machine learning / AI to achieve a specific objective.

Solution

A Data Study consists of the following phases:

  1. Data Survey
    We will first review and inspect the data available for use in the study. We will look at the volumes of data available, assess the quality and determine what is relevant. We will also prepare it for the next stage, for example, by performing some initial data cleansing, or by identifying whether part of the data needs to be further enriched or annotated.
  2. Data Exploration
    Our data science team will start to work with the data, training models on subsets and samples of the wider dataset to understand how best to draw out insights relating to the overall objective.
  3. Data Insight Report
    We will then deliver a report outlining: data overview; the techniques and approaches used; results achieved; and recommendations or considerations for delivering a future project.

Outcome

We have achieved good results in the Data Studies we have worked on for SITA to date. For example, we were able to predict ‘type of bag’ from an image taken at Bag Drop to 90% accuracy on the identified classes.

We therefore have positive indications that these projects are compatible for developing further towards a future PoC or full Production delivery.