Predicting Airbnb user's desired travel destinations

Abstract: The purpose of this report is to go through how to predict user’s intentions with machine learning and describe our thought process when working on the different stages of solving this type of problem. This will be done by solving Airbnb’s Kaggle problem where they wanted Kaggle users to predict where their users were most likely going to travel to based on data from their website. We go through the different choices we made while cleaning and prepairing the provided datasets and the reasoning behind these. The prediction is made using XGBoost and its boosted decision tree algorithm with several different approaches to how we go about preparing the data for training. Finally we upload the results for validation on the Kaggle challenge page and we discuss the strengths and weaknesses behind every approach and discuss what more could have been done to further improve the result.