A Smart City Application: Business Location Estimator using Machine Learning Techniques

Bilen T., Erel-Ozcevik M., Yaslan Y., Oktuğ S. F.

20th IEEE International Conference on High Performance Computing and Communications (HPCC) / 16th IEEE International Conference on Smart City (SmartCity) / 4th IEEE International Conference on Data Science and Systems (DSS), Exeter, Canada, 28 - 30 June 2018, pp.1314-1321 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/hpcc/smartcity/dss.2018.00219
  • City: Exeter
  • Country: Canada
  • Page Numbers: pp.1314-1321
  • Keywords: Smart Cities, Location Estimation, Relative Absolute Error, Regression, Hierarchical Clustering
  • Istanbul Technical University Affiliated: Yes


In smart cities, the business location is an important component of the complex decision process for the entrepreneurs. If a business such as shopping center, restaurant, coffee shop, hospital, clothing etc. is opened in optimal location; not only the satisfaction level of the customers are enhanced with increased life quality and convenience, but also it maximizes the profit of the entrepreneurs. However, it is complex and long-term decision to find optimal location for a business. Therefore; we propose a smart business application that uses machine learning techniques to estimate the location of a business. The proposed system collects key feature values for a specific business and learns a prediction model for future data. It estimates features and suggests clusters of districts that have optimal locations for that specific business. The proposed system is evaluated on a use case that estimates the best possible locations of a restaurant according to main characteristics taken as inputs from entrepreneurs via a Web based application. These characteristics are Meal Price per customer, HouseHold Type, Gender, and Age. The first phase of decision process is estimating mentioned features such as House Price per district and local population belonging to different HouseHold type, Gender, and Age with less error rates for the consecutive years. Here; we determine an optimal regression model choosing from SMORegression (SMOReg), MultiLayerPerceptron (MLP), and multivariate Linear Regression (Linear) according to Relative Absolute Error (RAE %) parameter, for each feature. The second phase clusters districts according to these estimated features by using hierarchical tree. As a result; the estimated best suggestions for restaurant locations which have similar features, are shown to entrepreneurs via our Web based application. It should be noted that, the application can produce generic solution for the needs of entrepreneurs.