Successfully reported this slideshow.
Your SlideShare is downloading. ×

Sorto-Project-Inls625-AirbnbListingCostandReviewPrediction.pdf

Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Loading in …3
×

Check these out next

1 of 18 Ad
Advertisement

More Related Content

Recently uploaded (20)

Advertisement

Sorto-Project-Inls625-AirbnbListingCostandReviewPrediction.pdf

  1. 1. Airbnb Listing Cost and Review Prediction Alex Sorto INLS625 November 2022
  2. 2. Problem Definition and Background Information • The motivation for this study is discovering and learning about the different machine learning techniques used by Airbnb when looking for a suitable listing. The process of selecting a listing could be automated and could save both money and time for the consumer. With this in mind, the study has the following two objectives: • Predict rental cost (price) for a specific number of people (beds, which are assumed 1 per person) • With a specific cost to pay (price), predict Airbnb reviews
  3. 3. Question 1 • Predict rental cost (price) for a specific number of people (beds, which are assumed 1 per person)
  4. 4. Methods • Step 1: Data Wrangling: Find and Replace all dollar signs in visual studio code
  5. 5. Problems Faced in the Project • Data dollar signs • Missing values in the dataset • Missing predicted values made it not possible to compute numeric scorer, so adjusting to same values means there is a margin of inaccuracy in the numeric scorer • RapidMiner did not understand excel data so I ran it in Knime • Low accuracy for predicting reviews with just price. Thus bedrooms were added but are different to the research question of the project • Decision Tree and SVM not enough java heap space, dataset had to be reduced to 10000 entries and SVM was still not possible due to limited heap space for both questions
  6. 6. Evaluation (Numeric Scorer) • Low R-squared, meaning more dispersion to predict price (Minitab Blog Editor ).
  7. 7. Prediction price vs Price of Airbnbs
  8. 8. Demonstration Open Street Map
  9. 9. Question 2 • With a specific cost to pay (price), predict Airbnb reviews
  10. 10. Methods
  11. 11. Evaluation Added bedrooms variable, which was not part of research question but increased accuracy by a small amount. Results of first trial with the only two variables
  12. 12. Review Scores Rating vs Predicted Reviews Line Plot
  13. 13. Demonstration Open Street Map
  14. 14. Part 3: Decision Tree Analysis for Both Questions (Same workflow)
  15. 15. Results for Predicting price
  16. 16. Results for Predicting Reviews
  17. 17. Conclusions • For predicted reviews, computed values similar to the mean of all reviews with little fluctuations. • Predicted price is more similar to what one would expect based on the line plot, and the behavior of both price and predicted price is similar. However, there were exceptions were there were peaks in price in some areas. • In conclusion, the first model may be useful for predicting price (with some exceptions of some areas that are more expensive), and the second model is useful to know the mean of reviews. • If we want more accurate predictions, decision tree analysis provides better results.
  18. 18. References • https://blog.minitab.com/en/adventures-in-statistics-2/regression-analysis- how-do-i-interpret-r-squared-and-assess-the-goodness-of-fit • http://insideairbnb.com/new-york-city

×