High performance computing for spatial scan statistics

Abstract

With the advent of mobile phones and wearable technology, there has been a sudden increase in the amount of geo-located data available to us. One of the basic forms of pattern recognition is the identification of hotspots. These can range from crime hotspots to disease-prone areas to identifying dense forest areas. In this report we implement a parallel version of Significant Linear Hotspot Discovery algorithm by using Neighbor Node filter optimization for detecting linear hotspots, a follow up on our previous work which was on circular hotspots.