Airbnb Analysis using Causal Inference in Machine Learning

Authors

AbdulRehman, Jerry Franklin , Sapna Sharma

Abstract

As more and more people prefer to stay at airbnb accommodation rather than staying at fancy hotels, the demand for airbnb has increased over time. Due to this demand more people are investing in houses to rent on Airbnb. Investors need to know which property to invest in to get high return on investment.

See Video abstract

How to explore this project

The main file is airbnb_analysis.ipynb notebook. It contains step by step implementation of causal inference on ROI of any investment listed on airbnb

Necessary api to collect the data are as follows :

1. Here Maps
2. Walkscore.com
3. greatschools.com

The data collected by our team is in the folder Data We have done the analysis and testing of the models in R and Python Languages. Analysis.ipynb is the python notebook and Global Markov & Failthfulness.ipynb is the code in R

Step 1: Data Gathereing

We filtered out data based on

• the property type (Condominiums, Apartments )
• only the properties which were available to rent as a whole

Step 2: Featue Engineering

We used Google's geo-coding API to get the addresses for the properties using their latitude and longitude.

Step3: Feature Engineering(Zestimate)

We used Zillow to get the Zestimates (current estimated market price) for each of these properties using their addresses.

The data from these two files were combined.

Step5: Model Building using Pyro

In the file airbnb_analysis.ipynb we build our causal models in pyro, analyze the results.

Directory Dictionary

├── home\n
│   ├── airbnb_analysis.ipynb   --- Main Notebook with the analysis
│   ├── Global Markov & Failthfulness.ipynb  		 --- R notebook with the tests used in the main notebook
│   
│   ├── listings.csv        			---data directly from Airbnb
│   ├── listings_manual.csv 			---listing with their price estimates, scrapped manually*
│   ├── listings_full.csv   			---Feature rich dataset with all collected features*
│