Airlines: Classified feedbacks into relevant categories and sent out hourly reports for assessment to the teams

Business Context

The engineering team of large south east Asian airline had to review each line of feedback to identify issue with air crafts

Key Questions

  • How are issues being categorized currently? What is the need for re classification?
  • What level of categorization with help teams? Single or multi level?
  • What proportion of issues categories? How can they identified?

Impact Created

  • Feedbacks could be categorized into desired issue category with 90% accuracy.
  • Previously classification of 200 issues took around 10-12 hours and it was reduced to 1-2 hours
  • A live report was generated for the engineering team to attend to the important classified problems

Solution Design

  • Categorization of already labelled data 4K feedbacks
  • Feature creation through nlp
  • Level 1(~10) classification using multiple classification techniques
  • Level 2(~5-6) classification using multiple classification techniques
  • Monthly runs along with improvements based on feedback of the model
Tags: Airlines | Operations