Price Prediction of Peer-to-Peer Accommodation Based on Sentiment Analysis and POI Feature Extraction

Abstract

In recent years, sharing economy is very popular around the world. Airbnb is the pioneer of sharing economy and the originator of peer-to-peer accommodation. This paper utilizes open source data on Airbnb website and Point-of-Interest data of Beijing to explore the influencing factors of housing rental. We perform sentiment analysis on textual reviews to obtain sentiment score and review features. By estimating the number of various Points of Interest around the house, we get geographical features. Price prediction of Airbnb houses has been done after extracting listing features, host features, review features and geographic features. After analysis, we use the Decision Tree (DT), the Support Vector Regressor (SVR), the Random Forest (RF) and other models to train the data, and finally achieve 76% of model accuracy of price prediction. There are three main contributions in this research. Firstly, the article proposes a new feature extraction method for Airbnb price prediction. It is verified on the data set that the method can effectively improve the model performance. Secondly, the article studies the influencing factors of Airbnb prices and explains the principles of smart pricing tools, which can provide hosts with improved strategies, helping them set more suitable prices and obtain greater benefits. Thirdly, the research results of this article are universal and suitable for online housing rental platforms, which can provide advice for the platform's recommended pricing and host's independent pricing.