首页|University of Montreal Reports Findings in Machine Learning (Spatial and spatiot emporal modelling of intra-urban ultrafine particles: A comparison of linear, no nlinear, regularized, and machine learning methods)
University of Montreal Reports Findings in Machine Learning (Spatial and spatiot emporal modelling of intra-urban ultrafine particles: A comparison of linear, no nlinear, regularized, and machine learning methods)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
New research on Machine Learning is th e subject of a report. According to news reporting originating from Montreal, Ca nada, by NewsRx correspondents, research stated, "Machine learning methods are p roposed to improve the predictions of ambient air pollution, yet few studies hav e compared ultrafine particles (UFP) models across a broad range of statistical and machine learning approaches, and only one compared spatiotemporal models. Mo st reported marginal differences between methods." Our news editors obtained a quote from the research from the University of Montr eal, "This limits our ability to draw conclusions about the best methods to mode l ambient UFPs. To compare the performance and predictions of statistical and ma chine learning methods used to model spatial and spatiotemporal ambient UFPs. Da ily and annual models were developed from UFP measurements from a year-long mobi le monitoring campaign in Quebec City, Canada, combined with 262 geospatial and six meteorological predictors. Various road segment lengths were considered (100 /300/500 m) for UFP data aggregation. Four statistical methods included linear, non-linear, and regularized regressions, whereas eight machine learning regressi ons utilized tree-based, neural networks, support vector, and kernel ridge algor ithms. Nested cross-validation was used for model training, hyperparameter tunin g and performance evaluation. Mean annual UFP concentrations was 13,335 particle s/cm. Machine learning outperformed statistical methods in predicting UFPs. Tree -based methods performed best across temporal scales and segment lengths, with X GBoost producing the overall best performing models (annual R = 0.78-0.86, RMSE = 2163-2169 particles/ cm; daily R = 0.47-0.48, RMSE = 8651-11,422 particles/cm). With 100 m segments, other annual models performed similarly well, but their pr ediction surfaces of annual mean UFP concentrations showed signs of overfitting. Spatial aggregation of monitoring data significantly impacted model performance . Longer segments yielded lower RMSE in all daily models and for annual statisti cal models, but not for annual machine learning models. The use of tree-based me thods significantly improved spatiotemporal predictions of UFP concentrations, a nd to a lesser extent annual concentrations."
MontrealCanadaNorth and Central Amer icaCyborgsEmerging TechnologiesMachine Learning
NETL
NSTL
