Prediction of Locations Using Unsupervised Learning Method to Open a Restaurant Branch

Abstract

One of the major factors that lead to a restaurant’s success is its location. This paper proposes a simple technique which uses K-Means Clustering, an unsupervised machine learning algorithm, to rank and predict the neighborhoods in a city to open new restaurant branches. Most important attributes like user rating, frequent visits and venue category are consider while clustering the locations. With the help of ‘beautiful soup’ package, Mapsofindia.com is scraped to identify the geographical locations of the city. Main components of master data are Postcode, Neighborhood name, Latitude and Longitude information of the city. After cleaning the data and clustering the locations, each and every cluster is carefully observed and it is found that Cluster number two has neighborhoods with ‘Indian Restaurant’ as most frequently visited venue. Geographical details of the locations are retrieved using geolocator python package and foursquare API is efficiently harnessed in identifying the place people love to visit in a specified location. This paper discusses how to predict the location for opening a new restaurant. The same methodology can be applied to predict a location for opening any consumer-based business like Pharmacy, Ice Cream shops, Bakery’s, boutique etc.