We use cookies to collect and share information. Read our privacy policy to learn more. You consent to our cookies usage if you continue to use this site.
Case study

Expedia Hotel Recommendations

Industry: Travel Region: USA Technology: Python Volume: 0.1 man years
The Challenge
When the number of accomodation options exceeds the ability to process them, choosing a hotel for your next vacation can become a challenge. With personalization being the main technology trend of the modern age, Expedia called for data scientists to help improve customer experience and boost business results.
Our Mission
Victor, a data scientist from the Sobolev Institute of Mathematics and a collaborator on Softaria projects, took part in the Kaggle competition, aiming to enable personalized recommendations for Expedia hotel searches. Since Expedia didn’t have customer-specific data to create a true personalization model, the participants were asked to predict the likelihood of a user staying at 100 hotel groups.
The Solution
The Expedia Hotel Recommendations competition featured 22,713 entries and 2,209 competitors. Using the three provided data files, Victor created a prediction model that took third place in the competition.
A Machine Learning Model to Predict Booking Outcomes for Expedia
Provided Data Description
Three data sets were provided: the training set from 2013 and 2014, the test set from 2015, the data set containing features extracted from hotel reviews.
The training and test datasets contained information on customer behavior, such as what customers searched for, whether they clicked or booked, and whether a search result was a travel package.
The booking outcome was defined as a hotel cluster. Hotel clusters were formed using Expedia’s in-house algorithm, which combined hotels with the same characteristics into one cluster, based on the hotels’ historical data.
The competitors were asked to predict booking outcomes for a user event, described by user search data and other attributes.
Want to know more about the project?

Read more case studies

See all of our previous works
Sorry, your files couldn't be uploaded. The upload mustn't exceed 10mb.
No file chosen
X
Thanks for contacting us.
We'll review and get back to you shortly.