stemflow: A Python Package for Adaptive Spatio-Temporal Exploratory Model
Yangkang Chen, Zhongru Gu, Xiangjiang Zhan · The Journal of Open Source Software · 2024
Stemflow is a user-friendly Python package for Adaptive Spatio-Temporal Exploratory Model (AdaSTEM (Fink et al., 2013)).AdaSTEM is a modeling framework that adopts "splitapply-combine" methodology (Wickham, 2011) -it adaptively splits data into spatiotemporal grids, train models for each grid, and combines the models for ensemble prediction.Models in stemflow follow the style of scikit-learn BaseEstimator class (Pedregosa et al., 2011).It provides one-line model creation, fitting, prediction, and evaluation.It implements spatio-temporal train-test-split and cross-validation functions.After model training, feature importance could be evaluated with spatio-temporal dynamics.Stemflow also provides functions for visualizing ensembles structured in model training and generating GIF file for predicted results to animate the spatio-temporal movement of animal population.