EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Comparative Evaluation of Machine Learning Models for Extreme Weather Event Classification in Los Angeles and the New Orleans Metropolitan Area

Zackary Miller

American Journal of Capstone Research · 2026

DOI: 10.70251/ye5wj337 ↗

Kurz gefasst

Fachartikel in der Zeitschrift American Journal of Capstone Research, 2026. Verfasst von Zackary Miller. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Extreme weather events can cause economic damage, infrastructure disruption and loss of life, making accurate prediction increasingly important. This study evaluated the ability of machine learning models to classify weather event categories using historical meteorological observations and storm event records from two geographic regions: Los Angeles and the New Orleans metropolitan area. Four machine learning algorithms, random forest, logistic regression, gradient boosting, and support vector machine (SVM), were trained using weather variables, including maximum temperature, minimum temperature, precipitation, wind speed, and month. For New Orleans, two weather representation approaches were investigated: a spatially averaged dataset and an expanded multi-location dataset containing observations from ten weather locations. To address class imbalance, storm events were grouped into two categories: Rain/flood and storm. Models were trained using an 80/20 train-test split, and models were further optimized through hyperparameter tuning. Performance was evaluated using accuracy and macro-averaged F1-score. Results showed that random forest and SVM consistently achieved the strongest performance across both study regions. In Los Angeles, the two best models, random forest and SVM, achieved macro F1-scores of approximately 0.66-0.68, while the New Orleans random forest, SVM, and gradient boosting models achieved substantially higher macro F1-scores of approximately 0.87-0.88. Performance differences between the averaged and expanded New Orleans weather representations were minimal, suggesting that spatially averaged weather data retained most of the predictive information needed for classification. Among the evaluated models, the performance of the Los Angeles gradient boosting model was an anomaly, achieving an accuracy of 0.88, lying inside the other tested models’ accuracy range of 0.81-0.94, and a storm F1 of 0.11, distant from the other tested models’ F1 range of 0.45-0.80. These findings demonstrate the potential of machine learning methods for weather-event classification and highlight the importance of geographic context, feature representation, and evaluation metrics in environmental prediction tasks.

Zitieren

APA

Miller, Z. (2026). Comparative Evaluation of Machine Learning Models for Extreme Weather Event Classification in Los Angeles and the New Orleans Metropolitan Area. American Journal of Capstone Research. https://doi.org/10.70251/ye5wj337

Vancouver

Miller Z. Comparative Evaluation of Machine Learning Models for Extreme Weather Event Classification in Los Angeles and the New Orleans Metropolitan Area. American Journal of Capstone Research. 2026. doi:10.70251/ye5wj337

BibTeX
@article{Miller2026,
  author = {Miller, Zackary},
  title = {Comparative Evaluation of Machine Learning Models for Extreme Weather Event Classification in Los Angeles and the New Orleans Metropolitan Area},
  journal = {American Journal of Capstone Research},
  year = {2026},
  doi = {10.70251/ye5wj337}
}
Kostenlos starten →