SDMtune

SDMtune optimizes species distribution models in R by providing data-driven variable selection, a genetic algorithm for hyperparameter tuning, support for artificial neural networks (ANN), boosted regression trees (BRT), maximum entropy modeling (MaxEnt), and random forest (RF), and evaluation using AUC-ROC, TSS, and AICc.

Key Features:

• Data-driven variable selection: Functions select predictors based on data-driven criteria to retain relevant environmental variables and reduce unnecessary complexity.
• Genetic algorithm for hyperparameter tuning: A genetic algorithm optimizes model hyperparameters to improve performance while mitigating overfitting.
• Model evaluation metrics: Model performance is assessed using area under the receiver operating characteristic curve (AUC-ROC), true skill statistic (TSS), and Akaike's information criterion corrected for small sample sizes (AICc).
• Support for multiple algorithms: Implements artificial neural networks (ANN), boosted regression trees (BRT), maximum entropy modeling (MaxEnt), and random forest (RF) for species distribution modeling.

Scientific Applications:

• Biodiversity conservation: Informing analyses that identify and predict species distributions relevant to conservation planning.
• Habitat management: Supporting habitat suitability assessments and management decisions based on modeled species-environment relationships.
• Climate change impact assessment: Evaluating potential shifts in species distributions under changing environmental conditions.
• Species response prediction and policy support: Enabling prediction of species responses to environmental change and providing evidence for policy decisions.

Methodology:

Implements data-driven variable selection, a genetic algorithm for hyperparameter optimization, and model evaluation using AUC-ROC, TSS, and AICc for models built with ANNs, BRTs, MaxEnt, and RF.