Machine-learning-aided prediction of cancer attributed mortality using natural radiation, major air pollutants, and temperature as influencing variables

Air pollution, radiation, and temperature have been linked with cancer mortality, but studies that used integrated machine learning and geographic information systems to predict it are limited. The aim of this study is to explore machine-learning models and geographic information systems to predict cancer-attributed mortality (2020–2022) using air pollutants, radiation, and surface air temperature as independent variables. Furthermore, the model efficiencies were validated with geospatial inputs as background.

The datasets were collected from the National Cancer Registry Programme of the Indian Council of Medical Research and the National Aeronautics and Space Administration. Major air pollutants such as nitrogen dioxide, formaldehyde, black carbon, sulfur dioxide, particulate matter, carbon monoxide, methane, ultraviolet and short wave radiation, and surface air temperature were analyzed to examine their effect on cancer-attributed mortality for the study period 2020–2022. Machine-learning models and geospatial tools were used in this study.

Carbon monoxide, ultraviolet radiation, particulate matter, and surface air temperature were associated with cancer deaths during 2020–2022. Notably, the extra trees regressor model performed well with R² values of 0.88 (2020), 0.83 (2021), and 0.83 (2022) respectively. A model validation framework was developed to evaluate prediction efficiencies when integrated machine learning and geospatial tools were used.

Generally, air pollutants and surface air temperature were associated with cancer-attributed mortality during the study period. This highlights the importance of machine learning and geospatial tools with proper model validation.