Author(s)

Yuze Gu

Corresponding Author

Yuze Gu

ABSTRACT

This study is aimed at building a predictive model for Airbnb prices, specifically focused on New York City. To ensure the accuracy of the model and the corresponding results, public governmental data was used in this study. All the records derived from the database were then randomly assigned into 2 groups: training sample (50%) and testing sample (50%). A linear regression model was built to predict the length of stay in a Medicare skilled nurse facility in 2015 using the training sample and then was applied in the testing sample for performance assessment. The results turned out to be stunning. Areas such as Bronx, Brooklyn, Queens, Staten were found out to be significantly less expensive than Manhattan. Private rooms, Shared rooms were less expensive than the whole home/house. These predictors can then be used to build the predictive model. Multiple R-squared was 0.5179 and the adjusted R-squared was 0.5176. The average mean squared error for the linear model in the testing sample was 0.21. The correlation between the predicted and the observed was 0.72. The min-max accuracy was 0.93. The mean absolute percentage deviation was 7.22%. In general, we identified the importance of predictors of Airbnb price in New York City, such as location, type of room, number of reviews, etc. Our predictive model suggested that areas such as Bronx, Brooklyn, Queens were among the lowest priced rooms and shared rooms were generally less costly than renting a whole apartment.

KEYWORDS

Economics, Airbnb ecosystem, hotels and accommodations, business, predictive model, artificial intelligence