Figure: Word cloud for the Fast Food cluster

Executive Summary

Implementing a clustering method of restaurants provides a way to create a data-driven approach of choosing where to eat. As such, this study sought to find out whether the said method will show significant results especially when done in the city of Makati. This study as well aimed to determine what the cluster’s features are and how it can help someone know where to satisfy their palettes. Data was gathered via the Zomato API, yielding 2375 data points pertaining to restaurants. Once mined and wrangled, the dataset was reduced via multiple factor analysis. After which, the cleaned data was processed via K-means clustering. Internal validation methods were used to determine the most appropriate number of clusters with the dataset. After implementing the K-means clustering method and internal validation checks, 4 clusters were chosen which were labeled as the following, namely: Quick Bites, Fast Food, Casual Dining and Full-Service Restaurants. The top features and top restaurants per cluster were taken to provide a brief idea of the themes of each cluster. This can help several stakeholders in providing an unsupervised learning method as a way of making informed decisions on where to go to eat.