Machine learning reveals global patterns of fish locomotion

This dataset supports a predictive model that infers propulsion patterns of ray-finned fishes (Actinopterygii) based on morphological traits. The model is implemented in Python and uses the Explainable Boosting Machine (EBM), an interpretable machine-learning method. The Python code for model training, evaluation, and prediction is available on: https://github.com/YuanFishLab/Ray-finned-fish-propulsion-pattern-predicted-model The dataset includes 17,034 morphological trait records representing 4,750 genera of ray-finned fishes. It also provides a movement-video dataset containing 2,966 video links covering 2,047 fish genera, which was used to assign or validate empirical propulsion patterns. The final propulsion-mode dataset reports the predicted propulsion patterns for all 4,750 genera. We have also uploaded example videos and links of 35 types of fish propulsion modes for reference. Additionally, we provide R scripts and corresponding data files used to analyze the geographic distribution, temperature distribution, and extinction-risk patterns associated with different propulsion modes. These resources allow users to reproduce the spatial and ecological analyses presented in the study. Together, the datasets and scripts offer a reproducible and interpretable framework for studying the diversity, evolution, and ecological correlates of propulsion strategies in ray-finned fishes.