Dynamic Network Model for Smart City Data-Loss Resilience Case Study: City-to-City Network for Crime Analytics.

Area of Science:

• Smart City Technologies
• Data Science
• Network Modeling

Background:

• Modern cities generate vast amounts of data from smart devices and sensors, enabling context-aware services.
• Smart city services rely on data streams, making them vulnerable to disruptions like data loss.
• Optimizing resource deployment (e.g., emergency services) is a key application of smart city data analytics.

Purpose of the Study:

• To present a dynamic network model designed to enhance the resilience of smart city services against data loss.
• To identify and utilize shared temporal trends in multivariate spatiotemporal data streams for improved data prediction.
• To demonstrate the model's ability to adapt to changing data flows and improve prediction robustness.

Main Methods:

• Developed a dynamic network model analyzing multivariate spatiotemporal data streams.
• Identified statistically significant shared temporal trends across different city data.
• Applied the model to city-based crime rate data from Montgomery County, MD, USA, comparing it with single-city auto-regression.

Main Results:

• The dynamic network model improved crime rate prediction and data loss robustness by leveraging inter-city temporal trends.
• A maximum performance improvement of 7.8% was observed for Silver Spring, with an average of 5.6% for high-crime cities.
• The model accurately identified optimal network connections based on prediction error minimization; city distance and weather were significant predictors.

Conclusions:

• The proposed dynamic network model significantly enhances smart city service resilience to data loss.
• Utilizing shared temporal trends across cities is more effective than single-city models for prediction and robustness.
• Geographic proximity and weather patterns are key factors influencing crime data trends and network connectivity.