Abstract: The Maximum Entropy (MaxEnt) model is widely used in ecological niche assessment, yet its performance depends highly on hyperparameter settings and is particularly prone to overfitting or underfitting in small-sample scenarios. To maximize modeling efficiency and predictive accuracy, this study used the island-endangered plant Neolitsea sericea as an example and adopted an adaptive genetic algorithm to optimize the hyperparameters of the MaxEnt model. The results were compared with those from grid search and standard genetic algorithms under the same modeling task. The results showed that the adaptive genetic algorithm achieved the best overall performance, with a regularization multiplier of 1.40. Compared with the grid search method, it significantly improved computational efficiency (time reduced by 97.30%) while maintaining better generalization ability (recall rate 0.82 vs. 0.75). In comparison with the standard genetic algorithm, it achieved a higher area under the receiver operating characteristic curve of 0.88 and reduced runtime by 27.18%. This method is suitable for island regions with limited samples and challenging field surveys, providing an effective technical reference for rapid habitat suitability assessment and conservation decision-making for endangered plants in island ecosystems.