Airbnb Predictive Price Recommender System
Industry
Hotel Industry
Skills
- Python
- SQL
- Streamlit
- HTML
- Heroku
Algorithms
- Linear Regression
- Decision Tree
- Random Forest
- XGBoost
- LIME
- SHAP
Introduction
Airbnb has stood the test of time and has proved to be a go to site for travel planning. With their newly introduced features of the Airbnb App, we can expect a lot of growth from the company. The dataset is taken from their official website insideairbnb.com. This provides a great platform to analyze the factors that contribute to the listings’ prices. The project objective is to help customers and hosts get a sense of Airbnb listing prices based on the important features that define the listing property. This is aimed at Airbnb too as they can optimize their search models and service better to their customers.
The Project covers analysis of the vital data set “Airbnb Business Case”. Visitors and hosts have been using Airbnb since 2008 to expand travel possibilities and give a more unique, personalised way of experiencing the world. Use LIME and SHAP for model interpretability and predictive modelling to develop market-specific forecasts with various variables. The data used in this research offers information on listing activity and indicators in the United States in 2019. It offers over 800,000 listings and 27 key characteristics, enough to learn more about hosts, geographical availability, and the analytics needed to make projections and draw conclusions.
Airbnb has established their market in accommodation business successfully since 2008 and has provided hosts and visitors with attractive options to visit places across the globe. Using the data collected by Airbnb for their listings, we have employed ML models that predict listing price based on user inputs over the UI. In this process we have used regression models like Linear Regression, Decision Tree, Random Forest, and XGBoost algorithms to calculate the price from the selective input variables and have used LIME and SHAP for model interpretability. For this project, we have limited the scope of the training data for United States listings only.