Airbnb predictive model

Working in close collaboration with my friend and esteemed data scientist, Najnin Sneha we were tasked with analyzing a sample of Airbnb’s data that covers a few rental locations within the District of Columbia. The data was collected in September of 2022 and includes data from neighborhoods within DC, such as Capitol Hill, Takoma, Fort McNair, as well as several variables that may influence demand, price, and rating of each rental.

Our goal was to analyze the data provided and build a predictive model for price of the listing. Upon creating multiple models, we can conclude that model 2, using the Caret nnet package gave us a better output of the predictive price.

Our team's analysis revealed that, despite removing certain features from the dataset and developing three neural network models, we achieved a prediction accuracy of around 50% for Airbnb prices in the DC area. This limitation might stem from the relatively small dataset made available by Airbnb. Neural networks typically require more extensive data compared to traditional machine learning algorithms to function optimally. Therefore, we suggest that Airbnb could enhance the dataset's utility by incorporating listings not just from DC, but also from Northern Virginia and Southern Maryland. This expansion could potentially increase the predictive accuracy of our neural network models.

For detailed insights and access to the source code and report, please visit the following GitHub link: