Spatial Data Mining for Optimized Selection of Facility Locations in Field-based Services

Spatial data mining has been developed as the effective technique in many applications that involve large amounts of geo-spatial data. Many organizations provide field-based services such as delivery, field-services and emergency to their customers. Considering the geographical distribution of the customer request points, the location of facilities will have noticeable impact on the overall efficiency of the company's operations. The closer the facilities are to the customers, the sooner and cheaper will be the service provision transaction. In this paper, we empirically study the role of spatial clustering methods in such context. We have implemented and tuned some of the main spatial clustering algorithms to discover the best locations for facility establishment. A new spatial clustering algorithm is proposed that does not require the number of facilities as input. The new algorithm will determine the optimal number of facilities along with their locations based on the business context trade-offs. Many experiments are conducted to study the performance of the studied algorithms on real world and synthetic data sets. The results reveal valuable distinctions between the different methods and confirm the higher efficiency of the proposed algorithm.