Predicting AsiaYo Users Spending for Improving Search Results - - PowerPoint PPT Presentation
Predicting AsiaYo Users Spending for Improving Search Results Travis Greene, Martin Hsia, Letitia She, Leo Lee Business Goal Stakeholders: Assumptions: AsiaYos managers 1. Model trained only on previous bookings 2. Avg. AsiaYo
Stakeholders: AsiaYo’s managers Assumptions: 1. Model trained only on previous bookings 2.
are price sensitive 3. All users are new users Challenge: 1. Plenty places to book a room for a trip 2.
Opportunity: Improve search results to increase conversion rate
Improving the default sorting (AY Sort) Reduce time for user to search and decide UX is improved by predicting customers’ budgets Better conversion rate
Predict the amount users will spend nightly Predictive and supervised task. We are taking past customers’ transaction data and predicting new users’ per night spending A predicted amount paid per night (numeric)
Guests
paid/night
paid/night Nights
day/month/time
Days of the Week
Platform User/Accom. Country Accom. City
History order transaction data 50,546 rows 16 columns 1. Remove internal test data, outliers, unnecessary rows 2. Bin time, Convert to day of week, keep months. Compute time differences between order creation & check-in date 3. Create new column #per_night Training ~40000 data rows, Testing ~10000 data rows 80/20 split
glm gbm
residuals_glm residuals_gbm
Reducing Model Prediction Error
search results
Algorithmic Considerations
trade-off
Business Policy
personal/demographic data