Jacho Saa,Jorge Josué(2023)MÉTODOS ESTADÍSTICOS BAYESIANOS APLICADOS EN LA MODELACIÓN DE LA DISTRIBUCIÓN POTENCIAL DE Fusarium oxysporum EN ECUADOR CONTINENTAL.Quevedo.UTEQ.92p.

Evaluating the efficiency of species distribution models is crucial for predicting the future distribution of a species. In this research, the efficiency of the Bayesian model, Bayesian Additive Regression Trees (BART), was evaluated against statistical models: Generalized Linear Model (GLM), Generalized Additive Model (GAM) and machine learning algorithms: Random Forest (RF), Maximum Entropy (MAXENT), Gradient Boosting Machine (GBM) to model the potential distribution of Fusarium oxysporum in continental Ecuador. Species occurrence records (GBIF) and bioclimatic variables (WorldClim) were used to build the models in RStudio at a spatial resolution of 10 x 10 km2. Models were evaluated using AUC, Miller calibration line and threshold-based classification. Presence probability maps from each model were exported from RStudio to ArcGis to generate binary maps and calculate the area of presence and absence of the species in Ecuador. The results indicate that the RF model had the best performance in most metrics, with threshold prevalence: Specificity (0.93), true skill statistic (TSS) (0.98); with optimal threshold: Specificity (0.98), TSS (0.99); for AUC (1). However, GLM presented good results in the Miller calibration line (1), and the largest area of phytopathogen presence (127200 km2) and the smallest RF (47100 km2). Although Bayesian models were efficient in predicting the distribution of F. oxysporum, as were the other methods, future research should be conducted with more occurrence records to improve accuracy.

• প্রধান লেখক: Jacho Saa , Jorge Josué (author)
• বিন্যাস: bachelorThesis
• ভাষা: spa
• প্রকাশিত: 2023
• বিষয়গুলি: Algoritmos de Aprendizaje Automático, enfermedad de Panamá, MAXENT, BART
• অনলাইন ব্যবহার করুন: https://repositorio.uteq.edu.ec/handle/43000/8096