Machine learning-based ensemble model predictions of outdoor ambient sound levels

Outdoor ambient sound levels can be predicted from machine learning-based models derived from geospatial and acoustic training data. To improve modeling robustness, median predicted sound levels have been calculated from an ensemble of tuned models from different supervised machine learning modeling classes. The ensemble is used to predict ambient sound levels throughout the contiguous United States. The training data set consists of 607 unique sites, where various acoustic metrics, such as overall daytime L₅₀ levels and one-third octave frequency band levels, have been obtained. Data for 117 geospatial features, which include metrics such as distance to the nearest road or airport, are used. The spread in the ensemble provides an estimate of the modeling accuracy. Results of an initial leave-one-out and leave-four-out validation study are presented.