Networks

Network based predictions

The state of the art in crime predication increasingly appears to be "network based". That is, looking at real street networks, and assigning risk to streets, rather than areal grid cells.

Techniques

• Hotspots on a network
• Scan statistics on a network

Case studies

• Case study Chicago

Technical details

• Geometry sources - Sources of road networks in the real-world (e.g. UK Ordnance survey etc.) from Open Data sources
• Into a graph - How to get data sources into our Graph class
• Geometry reduction - Some thoughts on speeding the algorithms up