Scribd
Upload
78 views2 pages

AirBNB Pricing Prediction

This document discusses features that could be used to predict AirBnB listing prices and group listings by neighborhood. It identifies text features, image features, and listing details as potential predictors. Features like amenities, location, reviews, and host response rate are highlighted. Text analysis techniques like bag-of-words modeling and sentiment analysis are proposed. Visual features could also be extracted from images. Supervised learning methods like KNN, linear regression, and Lasso/Ridge regression are recommended to build predictive models and group listings by neighborhood.

Uploaded by

navaneethan12
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
78 views2 pages

AirBNB Pricing Prediction

This document discusses features that could be used to predict AirBnB listing prices and group listings by neighborhood. It identifies text features, image features, and listing details as potential predictors. Features like amenities, location, reviews, and host response rate are highlighted. Text analysis techniques like bag-of-words modeling and sentiment analysis are proposed. Visual features could also be extracted from images. Supervised learning methods like KNN, linear regression, and Lasso/Ridge regression are recommended to build predictive models and group listings by neighborhood.

Uploaded by

navaneethan12
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

AirBNB Pricing Prediction

For AirBnB, there are two main options from where the features will be extracted:

1. Text features
2. Image features

Use a classifier method to group:

1. Neighborhood
2. Listing price based on Neighborhood

Determining Listing price:

Basic features from the neighborhood:

1. Prices
2. Amenities
3. Area of the house with # of beds/baths
4. Local Transit nearby
5. Host bio ??
6. Image thumbnail

Apart from the basic ones, more significant features could be:

1. Number of reviews
2. Host response rate
3. Number of references

Neighborhood Classification:

1. Crime rate
2. Population
3. Culture, night-life??
4. Facebook/Uber check-ins to see

Dynamic pricing tool consideration??

1. Already a tool called Aerosolve.  Can we come up with a substitute and not a complicated one

Dataset:

1. http://data.beta.nyc/dataset/inside-airbnb-data/resource/9d64399b-36d6-40a9-b0bb-
f26ae0d9c53f?view_id=33b9a800-4ed6-4d41-8f87-494c6c8582eb
2. For considering holiday seasons
http://data.beta.nyc/dataset/insideairbnbdata/resource/
ce0cbf4683f9414a8a1d7fd5321d83ca
3. For text analysis:
http://data.beta.nyc/dataset/insideairbnbdata/resource/8115833e8a0e4af68aed4d96a0ae0b73
Feature Extraction:

For pricing prediction, can we use only #1 or #2 can be helpful?


For neighborhood prediction, we might use 2,3,4 below:

1. Listing info Take the data and fill the missing values if needed. This info could be whether the
house is apartment, condo, dorm, # of bed, charge/guest/night

2. Bag of words  Info like Summary of listing, space, description, experiences offered  Which
technique can be used here?
Word-class: In paper, 9-word class were chosen people, nightlife, activities, style, accessibility,
culture, nature, amenities, and comfort – might not be needed

3. Text sentiment features  TextBlob package, which calculates the polarity of a segment of text
by averaging the polarity of each word in the text included in the package’s lexicon

4. Visual features  download all listing images and extract visual features.
Speeded Up Robust Features (SURF) descriptors from the 100 images using OpenCV

Analysis and Prediction:


1. KNN classifier can be used to neighborhood detection. We can remove few neighborhoods with
fewer listings. Determining optimal value of K will be important. We have always run K-classfier
and checked its R2 value with various value of K. According to paper, sklearn’s Recursive Feature
Elimination (RFE) was used
2. Regression Technique like Lasso or Ridge will be useful to find significant features while also on
bag of words to see which are good enough for deductions.
3. Train, Test using any of the CV method  Bootstrapping for better estimation of BETAs.
Also is R2, AIC, BIC only the measure for good fit?
4. As of now, polynomial or in general Linear regression technique but if any better algorithm is
taught in the class then we can use that.
5. Correlation between neighborhood and pricing will depend on the area of the house located.
How will we consider the intersection?

You might also like