M.S. Applied Data Science - Capstone Chronicles 2025

27

The results of this recommendation system were effectively visualized by marking them on an HTML map of the local geography which could easily be distributed digitally.

Conclusion Through this analysis, we demonstrated businesses can be ranked based on strength of physical cross-shopping relationships. This ranking reveals opportunities for strategic partnership at specific business locations and supports more effective prioritization of cross-selling efforts. Further, geographically filtered graph outputs better capture the spatial and consumer cross shopping patterns in physical locations. This aligns with how consumers actually navigate retail environments and in turn improves community detection discovery, evidenced by amplified modularity scores. Additionally, practical map-based visualizations produced by our model make outputs more interpretable and actionable for business users. By translating graph structures into more practical geographic representations, analysis becomes more accessible to non-technical stakeholders and directly supports decision making around partnership strategies and retail placement. Lastly, density of popular franchise locations (e.g. 7-Eleven, Starbucks) are observed through the geographic filtering analysis and validated on external sources such as Google Maps. Recommended Next Steps Future studies should expand this analysis by incorporating the temporal component available in the complete dataset. Doing so would capture expected seasonal fluctuations, uncover latent time-based behaviors between communities and further strengthen model performance. The complete dataset also presents the opportunity to explore other geographic regions. Replicating this analysis in different locations will help assess the generalizability of this study and validate its effectiveness in alternate market and geographic contexts.

235