This study explores the application of One-Class Support Vector Machines (OC-SVM) for predicting occurrences of Portuguese Manof-War (Physalia physalis) along the Brazilian coast. The model leverages daily wind data and citizen science sightings to identify environmental conditions associated with the presence of this species. Given the imbalance in the dataset — where only positive instances are available — we frame the problem as a one-class classification task. The proposed approach achieved over 70% accuracy and 64% F1 Score on two test sets, demonstrating its feasibility despite limitations in temporal and spatial feature representation. The study highlights the importance of data quality, sampling strategies, and feature extraction methods, suggesting directions for future research. These include incorporating seasonal and geospatial information and exploring alternative architectures. The findings contribute to the growing field of environmental monitoring using machine learning, especially for rare or underreported species.