Problem Solving with Data Science

As a Data Scientist at Sabre, I worked on projects that cut across the cross-section of Sabre’s business units, from airlines to hotels.

De-Duplication of Customer Profiles: For two leading global airlines, profile duplication is impacting their ability to customize and personalize offerings. Used fuzzy string matching and matching rules to identify multi-feature duplicated profiles.

Customer Segmentation: Completed customer segmentation for a leading luxury hotel chain to analyze their high value customers and their characteristics.

  1. Used K-means clustering algorithm to obtain a segmentation of customers at the hotel chain globally.
  2. Provided insights pertaining to regional behaviors of high value customers and spend trends.
  3. Presented results to the customer.

Rate Insights: Ongoing project for 2 major hotel chains to develop an advanced analytics tool that is expected to aid property audit managers decipher anomalies in rate and room configurations using anomaly detection.

  1. I delivered an initial prototype that led to a long-term partnership with one hotel chain and the chain is co-investing in the solution.
  2. Second phase uses ML clustering and classification techniques for anomaly detection to predict configuration anomalies that lead to revenue leakage for the chain for a major luxury chain that is losing in revenue due to incorrect price configurations and unbookable rooms. It then links the performance to productivity and makes inferences on bookability.

Name Matching using Fuzzy String Matching algorithms: Solution was delivered to three major airlines and was accepted for production.

  1. Accuracy of profile matches using algorithms that work across different languages and phonetics increased from 39% to 99.2%.
  2. Will improve bookability for loyalty customers on the airline leading to greater customer satisfaction.

Shopping Analytics: Proof of Concept for a shopping dashboard for hosted airlines

  1. provides the airline with a clear picture for shopping using a scoring model that uses external datasets as well
  2. Alerting on poorly converting markets

MVP accepted for production. Future iterations to include forecast of shops and bookings, competitive fare information and Performance metrics for airline markets.  Developed a Proof of Concept for both Revenue Management and Marketing persona.

Shopping Cache: Proof of Concept for analyzing how hosted airlines are coping with meta search engines and their bulk searches using a cache server.

  1. Dashboard designed on Tableau to analyze cached vs. non-cached shops and the scope to fine tune cache for improved performance
  2. MVP accepted for production and provided BA support to the team for production and helped create synthetic data for testing.