Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jul 2, 2026

Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy
07:13

Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy

Published on: February 25, 2021

Explainable and Optimized Gradient Boosting Algorithms for Near-Real-Time Prediction of Cyanobacterial Alert Levels

Marcelo A Cappelletti1,2, María Belén Sathicq3, M Julissa Atía2

  • 1Instituto de Investigaciones en Electrónica, Control y Procesamiento de Señales - LEICI (UNLP-CONICET), Facultad de Ingeniería, Universidad Nacional de La Plata, La Plata, Argentina.

Water Environment Research : a Research Publication of the Water Environment Federation
|July 1, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

PreserVamos: Ecological data on aquatic ecosystems in Argentina collected through citizen science monitoring.

Ecology·2026
Same author

Environmental degradation in the middle basin of Conchitas river, Buenos Aires: Two decades of urban and industrial pressure.

Environmental science and pollution research international·2026
Same author

Translocation of Epipelic Biofilms and Their Short-Term Responses to Urbanization Impacts in Nutrient Rich Streams.

Anais da Academia Brasileira de Ciencias·2021
See all related articles

This study developed a machine learning framework to predict harmful cyanobacterial blooms (HCBs) using physicochemical data. The model effectively addresses imbalanced data, improving early-warning systems for freshwater safety.

Area of Science:

  • Environmental Science
  • Data Science
  • Ecology

Background:

  • Harmful cyanobacterial blooms (HCBs) present significant risks to aquatic ecosystems, water resources, and public health.
  • Effective early-warning systems are crucial for mitigating the impacts of HCBs.
  • Predicting HCBs is challenging due to imbalanced data, where bloom events are rare.

Purpose of the Study:

  • To develop and validate a machine learning (ML) framework for predicting HCB alert levels.
  • To assess the impact of resampling strategies on model performance under imbalanced conditions.
  • To identify key environmental predictors of HCBs using interpretable ML.

Main Methods:

  • Four gradient boosting ML algorithms (LightGBM, XGBoost, etc.) were combined with 12 resampling techniques (SMOTE, etc.).
Keywords:
class imbalancecyanobacteriaearly‐warning systemsharmful algal bloomsmachine learning

Related Experiment Videos

Last Updated: Jul 2, 2026

Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy
07:13

Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy

Published on: February 25, 2021

  • A nested cross-validation framework ensured unbiased performance evaluation.
  • Model performance was assessed using metrics suitable for imbalanced data (recall, F1-score, balanced accuracy, MCC).
  • SHAP analysis was employed for model interpretability.

    • Main Results:

    • Resampling strategies, particularly SMOTE-based methods, significantly improved the detection of rare cyanobacterial bloom events.
    • LightGBM with SMOTE demonstrated superior recall and F1-score, indicating robust prediction of HCBs.
    • Water temperature, turbidity, and pH were identified as key predictors of HCB alert levels.
    • The framework utilizes routinely measured, low-cost sensor data for near real-time predictions.

    Conclusions:

    • Addressing class imbalance and employing rigorous validation are essential for accurate HCB prediction.
    • The proposed ML framework offers a practical and operationally feasible solution for HCB early-warning systems.
    • Interpretable ML enhances understanding of HCB drivers, supporting environmental management strategies.