# Data-Driven Malaria Prevalence Prediction in Large Densely-Populated Urban Holoendemic sub-Saharan West Africa: Harnessing Machine Learning Approaches and 22-years of Prospectively Collected Data

**Authors:** Biobele J. Brown, Alexander A. Przybylski, Petru Manescu, Fabio Caccioli, Gbeminiyi Oyinloye, Muna Elmi, Michael J. Shaw, Vijay Pawar, Remy Claveau, John Shawe-Taylor, Mandayam A. Srinivasan, Nathaniel K. Afolabi, Adebola E. Orimadegun, Wasiu A. Ajetunmobi, Francis Akinkunmi, Olayinka Kowobari, Kikelomo Osinusi, Felix O. Akinbami, Samuel Omokhodion, Wuraola A. Shokunbi  
**arXiv:** [1906.07502](https://arxiv.org/abs/1906.07502)  
**Date:** 2019-06-18  
**Summary page:** https://www.summarizepaper.com/en/paper/1906.07502v1/

## Key points

- Plasmodium falciparum malaria is a significant threat, causing over 200 million cases and half a million deaths annually worldwide.
- Majority of cases and deaths occur in sub-Saharan Africa, particularly among children.
- Researchers have developed LEMPS, a computational system that uses machine learning to predict malaria prevalence.
- LEMPS utilizes 22 years of data from Ibadan, a densely populated urban area in sub-Saharan West Africa with high transmission rates of malaria.
- LEMPS demonstrates strong prediction performance on previously unseen validation data.
- The system can handle multiple correlated features by balancing regularization strengths through L1-norm and L2-norm trade-offs.
- Data-driven systems like LEMPS enable regionally adaptable surveillance, effective management of control strategies, and optimized resource allocation in healthcare systems.

## Summary

Plasmodium falciparum malaria remains a significant threat to human life, causing over 200 million cases worldwide and resulting in half a million deaths annually. The majority of these cases and deaths occur in sub-Saharan Africa, particularly among children. To address this issue, researchers have developed and validated a computational system that leverages the predictive capabilities of machine learning approaches known as Locality-specific Elastic-Net based Malaria Prediction System (LEMPS). This system utilizes 22 years of prospectively collected data from Ibadan, a densely populated urban area in sub-Saharan West Africa with high transmission rates of holoendemic malaria. The dataset includes information on monthly prevalence as well as temporal, environmental, and host features from over 90,000 screened study participants who received clinical and community services between 1996 and 2017. LEMPS demonstrates strong generalization performance in terms of both magnitude and direction when predicting monthly prevalence on previously unseen validation data with mean absolute error (MAE) of less than or equal to 6x10-2 and mean squared error (MSE) of less than or equal to 7x10-3 within an error tolerance range of +0.1 to -0.05. These prediction accuracies are relevant for supporting decision-making in holoendemic settings due to its ability to handle multiple correlated features by balancing regularization strengths through L1-norm and L2-norm trade-offs. Data-driven systems like LEMPS play a critical role in enabling regionally adaptable surveillance, facilitating effective management of control strategies, and optimizing resource allocation within stretched healthcare systems. In conclusion, the refined summary highlights the development and validation of LEMPS which has the potential to significantly improve malaria surveillance, control strategies, and resource allocation in highly endemic regions.