Czy AI może przewidywać przyszłe punkty zapalne przestępczości w mieście, analizując obrazy satelitarne i dane spisowe ?

Systemy uczenia maszynowego obecnie łączą obrazy satelitarne, trendy demograficzne oraz historyczne zapisy przestępczości, aby przewidywać, gdzie w nadchodzących miesiącach może wzrosnąć liczba określonych przestępstw. Te prognozy są wykorzystywane przez niektóre miejskie programy bezpieczeństwa do alokacji zasobów.

Background

Researchers have made significant progress in using machine learning algorithms to analyze satellite imagery and census data for predicting crime hotspots. By leveraging satellite imagery, AI models identify environmental factors such as urban decay, poverty, and lack of green spaces that are associated with higher crime rates. Census data provides additional insights into demographic and socioeconomic factors that can contribute to crime. Studies have shown that combining these data sources can improve the accuracy of crime predictions. For instance, a model that analyzes satellite images to identify features such as abandoned buildings, poor lighting, and dense vegetation can be combined with census data on population density, income levels, and education to predict areas with high crime rates. While this approach shows promise, its effectiveness can vary depending on the quality of the data, the specific algorithms used, and the local context. Furthermore, there are concerns about potential biases in the data and the risk of perpetuating existing social inequalities. The development of more sophisticated and nuanced models that can account for these complexities is an active area of research.

+- administered May 13, 2026 · Source: National Institute of Justice — Science Direct