Project ID: 201912-16
Team Members:
- Zihe Wang ([email protected])
- Qianhui Yu ([email protected])
- Duanyue Yun ([email protected])
Our project explores the importance of listing name and builds a web application providing visualization tools and prediction tools both for Airbnb hosts and guests. For visualization tools, we provide important words recommendation, market information delivery such as price heat map. For prediction tools, hosts can get estimated popularity of their listings by inputting their listing name and other listing information. We tried with two modeling methods (Random Forest and XGBoost) to predict popularity and compare them to a baseline model. We find that listing names indeed help predict popularity and rank high in terms of variable importance.
airbnb_scrape
contains the Web Crawler App we developed.airbnb_analysis
contains the Django App we developed.exploratory analysis
contains code for exploratory analysis.modeling
contains code for predictive model experiement.
-
pip install -r requirements.txt
to install required packages. -
For scrapy, run
scrapy crawl airbnb -a min='410' city='New York'
to get the data of your selected city and minimal price. -
For website, run
python manage.py runserver
to access it on localhost.
hw0-252420.appspot.com
The code developed by Qianhui and Duanyue are marked with UIN. The codes are developed by Zihe otherwise.
[1] R. Martinez, A. Carrington, T. Kuo, L. Tarhuni, and N. Abdel-Motaal. The impact of an airbnb host’s list-ing description ‘sentiment’ and length on occupancy rates.https://arxiv.org/pdf/1711.09196.pdf
[2] J. Pennington, R. Socher, and C. Manning. Glove: Globalvectors for word representation, 2014
[3] https://towardsdatascience.com/creating-word-clouds-with-python-f2077c8de5cc