Collaborating Authors


Machine learning helps to predict new violence hot spots


Violence reduction strategies, which have traditionally focused on pubs and nightclubs, need to be broadened to include places where alcohol is not served if they are to be effective, a new study has suggested. Researchers from Cardiff University's Violence Research Group gathered data from 10 city centers across England and Wales and used machine learning to map the distribution of reported incidents of violent crime against alcohol outlets and, crucially, locations where alcohol is not sold. When they compared their analysis to a model mapping only places where alcohol is sold, the researchers discovered their new combined model more accurately predicted levels of violence. The study found that alongside pubs and bars several other destinations often associated with typical "nights out" were hot spots for violent crime, including fast-food outlets, takeaways, bus stops and cash machines. The researchers say this is the first time such a wide area has been analyzed and their study has uncovered previously unmapped violence hot spots.

Police facial recognition camera trial hindered by software setbacks

Daily Mail - Science & tech

A police force's use of facial recognition technology requires'considerable investment' to deliver consistent results, a study has concluded. Crashing computer systems and poor quality images are among the challenges South Wales Police officers have faced since rolling out the technology. Large crowds, low lighting and people wearing glasses were all issues the AI software struggles to cope with, experts found. South Wales Police force first deployed automated facial recognition at the 2017 Champions League final in Cardiff. This led to the technology wrongly matching more than 2,000 people to possible criminals.