As nations worldwide commit to carbon neutrality targets in response to accelerating climate change, the spatial modeling of carbon emissions has emerged as an indispensable tool for policy implementation and assessment. This paper presents a systematic review of the field from bibliometric and methodological perspectives. We synthesize key developments in spatial allocation techniques, data-driven models, and emission characterization methods. A central focus is the transformative role of geospatial big data in improving model accuracy and applicability, particularly how fine-grained, high-resolution modeling enhances the efficacy of emission reduction strategies. Our analysis reveals several key conclusions. First, the literature on carbon emission spatial modeling is expanding rapidly, with a discernible shift in focus from coarse, large-scale assessments toward more granular analyses that are sector-specific, high-resolution, and multidimensional. Second, hybrid models that integrate top-down and bottom-up approaches are now the predominant strategy for enhancing both accuracy and applicability; coupling mechanistic models with machine learning techniques effectively reconcile macro-scale data consistency with micro-scale heterogeneity. Third, the integration of geospatial big data is revolutionizing the field by providing the high-resolution, multidimensional, and dynamic inputs necessary to transition from macro- to micro-scale analysis. This is particularly evident in fine-grained assessments of urban systems—including spatial functions, morphology, and transportation networks—where such data dramatically improve the characterization of emission sources, intensities, and their spatiotemporal heterogeneity. This study ultimately elucidates the critical role of fine-grained modeling in advancing the quantitative understanding of carbon emission drivers, enabling robust scenario simulations for carbon neutrality, and informing effective low-carbon spatial planning. The synthesis presented here aims to provide a firm theoretical and technical foundation to support the ambitious carbon reduction targets set by nations worldwide.

Multiple-instance learning (MIL) can solve supervised learning tasks, where only a bag of multiple instances is labeled, instead of a single instance. It is considerably important to develop effective and efficient MIL algorithms, because real-world datasets usually contain large instances. Known for its good generalization performance, MIL based on extreme learning machines (ELM–MIL) has proven to be more efficient than several typical MIL classification methods. ELM–MIL selects the most qualified instances from each bag through a single hidden layer feedforward network (SLFN) and trains modified ELM models to update the output weights. This learning approach often performs susceptible to the number of hidden nodes and can easily suffer from over-fitting problem. Using Bayesian inferences, this study introduces a Bayesian ELM (BELM)-based MIL algorithm (BELM–MIL) to address MIL classification problems. First, weight self-learning method based on a Bayesian network is applied to determine the weights of instance features. The most qualified instances are then selected from each bag to represent the bag. Second, BELM can improve the classification model via regularization of automatic estimations to reduce possible over-fitting during the calibration process. Experiments and comparisons are conducted with several competing algorithms on Musk datasets, images datasets, and inductive logic programming datasets. Superior classification accuracy and performance are demonstrated by BELM–MIL.