Counting 3,319 Big Data & Machine Learning Frameworks, Toolsets, and Examples...
Suggestion? Feedback? Tweet @stkim1


MG-GY-9753 | Business Analytics | AirBnB Country Prediction



Airbnb is a community marketplace where guests can book living accommodations from a list of verified hosts. Membership to the site is completely free and there is no cost to post a listing. Using a targeted user interface designed to narrow down traveling preferences, Airbnb offers an attractive, cost-saving alternative to traditional hotel bookings and vacation home rentals.

Upon finding a desired listing, guests are prompted to sign up for membership, which provides access to contact the host directly as well as provide payment information for a request. Only once the host accepts the transaction and the guest checks in is the credit card charged, along with a 6-12% transaction fee from Airbnb.

The process is similarly simple for hosts, who 3 receive a notification once a guest indicates interest in a particular listing and have the option to approve or deny the transaction. Once the listing is booked, the host receives the payment and Airbnb takes a 3% transaction fee.

The Challenge

Design a model that will predict where will a new guest book their first travel experience. By accurately predicting where a new user will book their first travel experience, Airbnb can share more personalized content with their community, decrease the average time to first booking, and better forecast demand.

In building this model, you will need to consider the following:

● Data size: Some datasets are big (more than 1M variables). Consider reducing the dataset to a random sample of 10% to run experiments. In addition, you will need to join datasets using “user id”

● Data quality: Some numerical and categorical values are missing. You can replace missing values with the media or mean. Alternative, you can eliminate the missing values by subsetting your dataset. Both approaches have tradeoff.

● Multiclass Model: To solve this, you need to build a “multi-class classification” model. This means you will need several classification to predict countries. Some approaches may be:

  1. Break up your problem into one classifier for new booking followed by classifiers by each destination

  2. Use Random Forest

  3. Use the CARET package in R

Pictorial Data Description



The Approach