Price Classification for NYC Airbnb listings

The objectives for this project were to predict the price category of Airbnbs in New York City in order to assist hosts when they price their Airbnbs.

I classified Airbnbs into two categories, budget-friendly or high-end, based on features such as neighborhood, room type, availability during the year, and frequency of reviews. To do this, I utilized five different machine learning methods: k-NN, Naive Bayes, SVM, Random Forest, and Logistic Regression.

After reviewing the confusion matrix, accuracy, and AUC scores from each model, I found that, overall, Random Forest provided the best fitting model for classifying both Airbnb price categories. I can also note that k-NN and SVM Linear performed well during high-end classification according to the AUC and F1 Score metrics.

However, ultimately, because I believe Airbnb consumers to be price sensitive, I recommend Random Forest as the best option for classifying price categories because it tends to generate lower prices than other algorithms and will be more attractive to consumers. I will use this algorithm to implement a Price Recommendation System for Airbnb hosts to use. Airbnb hosts will be able to input features of their rental and receive a recommended price category for their listing: budget-friendly, mid-range, or high-end.

For future analysis, to improve predictive accuracy, we can identify more variables that affect pricing such as square footage and additional services Airbnb hosts may provide.