Comparative Analysis of Machine Learning and AHP Models for Cafe Location Suitability Using Geospatial Data
Comparative Analysis of Machine Learning and AHP Models for Cafe Location Suitability Using Geospatial Data
Authors:
Chakka Nandita
Department of CSE - Data Science
Geethanjali College of Engineering and Technology
Hyderabad, India nandita.chakka@gmail.com
Manoor Mounika
Department of CSE - Data Science
Geethanjali College of Engineering and Technology
Hyderabad, India mounikamanoor26@gmail.com
Thiramdas Akshay Kumar
Department of CSE - Data Science
Geethanjali College of Engineering and Technology
Hyderabad, India thiramdasakshay.333@gmail.com
SRIHARI BABU GOLE
Department of CSE - Data Science
Geethanjali College of Engineering and Technology
Hyderabad, India sriharibabug@gmail.com
Abstract—The selection of the location is a crucial factor in the establishment of a successful cafe business. Traditionally, the site selection for a location is done based on the owner’s intuition. This paper presents a comparative analysis between predictive models using machine learning and analytic models representing an individuals intuition at the time of decision making. A self constructed Hyderabad cafes dataset was built using Google´ Places API and OpenStreetMap data, incorporating features related to competition, accessibility, and urban infrastructure.
Two tree based ensemble learning models, Random Forest and XGBoost, were trained to predict a success index derived from ratings and customer engagement. An Analytical Hierarchy Process (AHP) model was developed to provide an interpretable benchmark for comparison with the predictive models. The experimental results showed that the XGBoost model achieved the best performance with an R2 score of 0.828, outperforming the Random Forest model with an R2 score of 0.799, while the AHP model showed lower predictive performance with an R2 score of 0.309. Cross validation further confirmed the robustness of the XGBoost model with an average R2 score of 0.909.
These results highlight the effectiveness of machine learning based approaches in identifying high potential locations, emphasizing the role of accessibility, competition density, and urban features in determining business success. The proposed framework can help entrepreneurs and urban planners make informed and data-driven decisions about site selection.
Index Terms—Location suitability, cafe business, machine learning, XGBoost, random forest, geospatial analysis, site selection