AirBnB price prediction in USA

Why is this an intersting project?

• Pricing is one of the most important decisions that hosts need to make to gain profits.
• The prediction exercise here provides a method to predict prices for properties around the same area with similar conditions.
• It will give the hosts a better idea of the property’s market values.
• Airbnb can retain hosts, expand its business, and eventually increase revenues.

Data Source : https://www.kaggle.com/datasets/kritikseth/us-airbnb-open-data

Methodology

• Data preprocessing: Drop missing data and remove outliers
• Exploratoy Data Analysis:
  • Create histograms and correlation matrix for numeric features
  • Visualize listings and prices distributions across states
• Data Modeling:
  • Use linear regression, XGBoost and multivariate adaptive regression splines to predict prices
  • Compare the model performances by RMSE, MAE and R2 Score

Outcomes

• Price is most positively correlated with host listings count and availabilty in the year.
• Price is correlated with Room types and city.
• So, more listing a host has, the price is higher and if the place is more available, you can charge more!

Looks like renting a private room is more attractive over other options!

Conclusion

• Predicted prices with three models: linear regression, multivariate adaptive regression splines, and XGBoost.
• XGBoost has the best performance with the highest R square and lowest RMSE and MAE.
• Room types and city are the key predictors.