• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Thursday, November 19, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Technology

Machine Learning Accurately Forecasts Monthly Malaria Incidence Across Southeastern Senegal Districts

by
November 9, 2026
in Technology
Reading Time: 5 mins read
0
Machine Learning Accurately Forecasts Monthly Malaria Incidence Across Southeastern Senegal Districts

Machine Learning Accurately Forecasts Monthly Malaria Incidence Across Southeastern Senegal Districts

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

In Senegal, malaria transmission remains highly heterogeneous, concentrating severe health burdens within the southeastern regions of Kolda, Kédougou, and Tambacounda, collectively known as the KKT zone. Although these three administrative regions account for merely 11.3 percent of the national population, they persistently endure the vast majority of disease transmission, registering 78.5 percent of all reported malaria cases and 43.6 percent of malaria-related fatalities in 2021. To counter this endemicity, Senegal’s National Malaria Control Program has deployed extensive clinical and vector interventions, encompassing rapid diagnostic tests, artemisinin-based combination therapies, insecticide-treated nets, intermittent preventive treatment during pregnancy, and seasonal chemoprevention. Nevertheless, annual incidence rates in the KKT territory have consistently exceeded 300 cases per 1,000 inhabitants, highlighting the urgent need for predictive tools to optimize intervention deployment.

The complex epidemiology of southeastern Senegal is deeply shaped by bioclimatic variation that dictates mosquito vector ecology and transmission seasonality. Senegal encompasses seven distinct bioclimatic zones, creating an environmental gradient ranging from arid Sahelian conditions in the north to humid Sudano-Guinean climates in the south. Annual rainfall reflects this geographic division, varying from approximately 300 millimeters in northern areas to over 1,200 millimeters across southern districts. Precipitation follows a strict seasonal cycle, dividing the year into a dry season and an intense wet season extending from May through November, with rainfall peaking during July, August, and September. This climatic rhythm drives vector breeding habitats, keeping transmission low during dry months before triggering steep infection spikes once rains commence. Multi-year records indicate that this regional disparity has persisted over time, with the KKT zone accounting for 82 percent of national cases in 2018, 81 percent in 2019, 83.3 percent in 2020, and 78.5 percent in 2021.

To construct a comprehensive predictive system, the researchers assembled a longitudinal database covering thirteen health districts across the KKT zone between January 2016 and December 2021. The study area encompassed three districts in Kolda, specifically Kolda, Médina Yoro Foulah, and Vélingara; three districts in Kédougou, including Kédougou, Saraya, and Salémata; and seven districts in Tambacounda, covering Bakel, Dianké Makha, Goudiry, Kidira, Koumpentoum, Makacolibantang, and Tambacounda. The final dataset included 936 monthly district-level records structured across 17 variables, with no missing entries. Clinical and intervention metrics provided by the National Malaria Control Program supplied confirmed monthly case numbers, quantities of artemisinin-based combination treatments distributed across age categories, coverage of intermittent preventive treatment in pregnant women, and binary tracking of seasonal chemoprevention. Meteorological parameters obtained from the National Aeronautics and Space Administration provided monthly measures of temperature, precipitation, humidity, soil moisture, and solar radiation, while confirmed cases were lagged by twelve months to capture temporal autocorrelation.

The research team evaluated three gradient boosting algorithms to forecast monthly malaria incidence: Extreme Gradient Boosting, Histogram Gradient Boosting, and the Light Gradient Boosting Machine. These three ensemble architectures were configured to predict monthly incidence per 1,000 people across the thirteen districts by capturing non-linear relationships among environmental conditions, past caseloads, and public health interventions.

The 936 observations were partitioned chronologically, using data from January 2016 through December 2019 for training, 2020 for testing, and 2021 for independent validation, while 2022 was reserved for prospective forecasting. Hyperparameters were tuned using Bayesian optimization paired with five-split TimeSeriesSplit cross-validation on the training set. For Histogram Gradient Boosting, optimization identified an optimal tree depth of 7.26, a learning rate of 0.08, 462.79 iterations, 196.88 maximum bins, and beta regularization of 0.84. Light Gradient Boosting converged on a maximum depth of 10, a learning rate of 0.3, 301.59 estimators, and a column subsampling ratio of 0.7. Extreme Gradient Boosting achieved optimal configuration with a depth of 3.66, a learning rate of 0.02, 387.97 estimators, a column fraction of 0.61, and alpha regularization of 0.03, alongside bootstrapping with 200 resamples to evaluate uncertainty.

Model performance was evaluated using mean absolute error, root mean square error, and the Nash-Sutcliffe coefficient of determination. In temporal cross-validation on training data, Extreme Gradient Boosting demonstrated the best performance, attaining a mean root mean square error of 7.59 per 1,000 with a 95 percent confidence interval of 3.14 to 12.04 per 1,000. In comparison, Histogram Gradient Boosting achieved a mean error of 9.16 per 1,000 with a confidence interval of 5.03 to 13.28 per 1,000, while Light Gradient Boosting recorded 9.52 per 1,000 with an interval of 6.87 to 12.16 per 1,000. On the 2020 test set, Extreme Gradient Boosting maintained superior performance, yielding the lowest root mean square error of 12.74 per 1,000, the lowest mean absolute error of 6.38 per 1,000, and the highest coefficient of determination at 0.90, compared to 13.87 per 1,000 and 0.88 for Histogram Gradient Boosting, and 13.73 per 1,000 and 0.89 for Light Gradient Boosting.

Validation on the independent 2021 dataset confirmed Extreme Gradient Boosting as the most accurate model for malaria incidence forecasting across southeastern Senegal. On this final validation benchmark, Extreme Gradient Boosting achieved its highest overall accuracy, recording a coefficient of determination of 0.93 alongside a root mean square error of 10.58 per 1,000 and a mean absolute error of 5.95 per 1,000. In comparison, Histogram Gradient Boosting produced a root mean square error of 11.93 per 1,000, a mean absolute error of 6.16 per 1,000, and a determination coefficient of 0.91, while Light Gradient Boosting registered 12.30 per 1,000, 6.96 per 1,000, and 0.91, respectively. Examination of prediction intervals indicated that the 95 percent confidence intervals were narrow and stable on test data, reflecting good reliability even during transmission peaks. Although validation intervals widened across certain high-incidence districts, the model demonstrated excellent local precision, generating forecasts that tracked observed values very closely in the districts of Kidira, Vélingara, and Koumpentoum.

Prospective forecasts generated by Extreme Gradient Boosting for 2022 captured nuanced spatiotemporal variations in malaria incidence across the thirteen districts of the KKT zone. Initial incidence increases appeared as early as June across five districts: Bakel, Kédougou, Médina Yoro Foulah, Salémata, and Saraya. By July, transmission increases broadened to encompass Dianké Makha, Goudiry, Kolda, Koumpentoum, Makacolibantang, Tambacounda, and Vélingara. Furthermore, the model predicted distinct timings for peak malaria incidence. Salémata and Saraya reached peak transmission exceptionally early in August. Conversely, seven districts—Bakel, Dianké Makha, Goudiry, Kidira, Koumpentoum, Kédougou, and Tambacounda—reached peak incidence in October. In contrast, Kolda, Makacolibantang, Médina Yoro Foulah, and Vélingara peaked later in November. With the exception of Salémata and Saraya, these predicted seasonal peaks showed strong concordance with previous epidemiological studies by Diouf and colleagues, demonstrating the model’s ability to delineate localized transmission timing for public health planning.

To interpret the internal mechanics of Extreme Gradient Boosting, researchers utilized Shapley additive explanations derived from cooperative game theory to quantify the marginal contributions of explanatory variables. The analysis identified confirmed malaria cases lagged by twelve months as a highly influential factor, reflecting persistent transmission cycles in historically affected areas. Elevated soil moisture and relative air humidity consistently increased predicted incidence, pointing to environmental conditions conducive to mosquito breeding, although no causal links can be established from these associations. Population density also exhibited strong influence, with low densities generally associated with higher predicted incidence and higher densities linked to lower predictions. Seasonal chemoprevention implementation was typically associated with reduced predicted incidence, while temperature and rainfall displayed non-linear patterns, where moderate temperatures and adequate rainfall increased predictions, but extreme heat or excessive deluges lowered them, and solar radiation showed negligible direct effect.

To translate these predictive capabilities into practical public health interventions, the authors developed PrevAPP, a web application powered by the optimal Extreme Gradient Boosting model. PrevAPP enables health authorities to load multi-source epidemiological and meteorological datasets and generate prospective monthly forecasts across southeastern Senegal with a twelve-month horizon, including filtering capabilities by specific months. Despite these encouraging results, the researchers acknowledged several study limitations. The dataset comprised 936 observations across six years, representing a moderate sample size that may constrain broader generalizability. Furthermore, satellite climatic data can suffer from spatial resolution limits, while clinical records may be subject to underreporting, diagnostic errors, or reporting inconsistencies. The modeling framework also lacked data on human behavior, population migration, local healthcare infrastructure quality, and long-lasting insecticidal net coverage. Future research directions include incorporating population mobility and bed net metrics, exploring recurrent neural networks and attention-based models, and conducting prospective real-world pilot trials to optimize malaria management across endemic settings.

Subject of Research: Machine learning forecasting of monthly malaria incidence in southeastern Senegal

Article Title: Forecasting malaria incidence in the Southeast districts of Senegal using a machine learning approach

Article References: Ly, A. Y., Allaya, M. M., Loum, M. A., Gning, L. D., Sy, O., Bakhoum, L. N., Ndiaye, K., Ndiop, M., Sall, F. B., Ndiaye, J. L. A., & Bousso, M. (2025). Forecasting malaria incidence in the Southeast districts of Senegal using a machine learning approach. BMC Artificial Intelligence, 1(1), Article 9. https://doi.org/10.1186/s44398-025-00007-4

Image Credits: AI Generated

DOI: 10.1186/s44398-025-00007-4

Keywords: malaria forecasting, machine learning, XGBoost, Senegal, epidemiology, public health, disease surveillance, climate data, Forecasting, malaria, incidence, Southeast

News Source: Teresa Odom. (November 9, 2026). Machine Learning Accurately Forecasts Monthly Malaria Incidence Across Southeastern Senegal Districts. Scienmag.

Tags: climate datadisease surveillanceEpidemiologyforecastingincidenceMachine Learningmalariamalaria forecastingPublic HealthSenegalSoutheastXGBoost
Share12Tweet7Share2ShareShareShare1

Related Posts

Researchers Construct Five-Dimensional Spinning Black Holes Dual to Plasma Balls

Researchers Construct Five-Dimensional Spinning Black Holes Dual to Plasma Balls

November 19, 2026
Lithium Slag Strengthens Stabilized Clay Road Bases and Cuts Carbon Emissions

Lithium Slag Strengthens Stabilized Clay Road Bases and Cuts Carbon Emissions

November 19, 2026

New Matrix Method Improves Vibration Accuracy for Restrained Orthotropic Plates

November 19, 2026

First-Principles Calculations Reveal High Thermoelectric Efficiency and Half-Metallicity in Co2VPb Alloy

November 19, 2026

POPULAR NEWS

  • First-Principles Calculations Reveal High Thermoelectric Efficiency and Half-Metallicity in Co2VPb Alloy

    First-Principles Calculations Reveal High Thermoelectric Efficiency and Half-Metallicity in Co2VPb Alloy

    29 shares
    Share 12 Tweet 7
  • Mobile Bird Hosts Enable Regional Gene Flow in Crater Lake Parasites

    29 shares
    Share 12 Tweet 7
  • Ultrasound Shows Low Diagnostic Accuracy for Sexing Immature Agile Frogs

    29 shares
    Share 12 Tweet 7
  • Lithuanian Forest Foragers Link Solitary Mushroom Gathering to Cross-Border Trade

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Structured Omega-3 Supplementation Linked to Favorable Whole-Blood Biomarkers in Large Study

Delaying Surgery Up to Nine Hours Proves Safe in Complicated Appendicitis

Cooling Paradoxically Boosts Cav3.2 Calcium Influx During High-Frequency Neuronal Firing

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 88 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.