Scholarlink

Typhoid fever remains a significant public health challenge in Africa particularly in regions with limited access to clean water and sanitation. The financial burden associated with typhoid fever treatment can exacerbate disparities in healthcare access and affordability, especially in resource-constrained settings. By leveraging data analytics and machine learning techniques, this paper seeks to develop predictive models capable of accurately assessing the risk of typhoid fever and guiding resource allocation strategies. Through the integration of diverse datasets, including clinical parameters, demographic information, and environmental factors, the proposed modeling framework aims to identify key predictors of typhoid fever incidence and severity. The dataset comprises laboratory tests of blood samples of one hundred and fifty (150) children, from age zero (0) to five (5) years. Data preprocessing and training were performed using Waikato Environment for Knowledge Analysis (WEKA) workbench tool as a machine learning platform. The information Gain (IG) feature selection technique was employed to select the best attributes using a cut-off gain of 0.2. The top six (6) attributes with more information gain relevant to the class typhoid were selected. Widal test, Malaria count, Monocyte, Platelets, HB, Eosinophils, cost of treatment are the selected attributes with their respective ranked information gain of 0.6759, 0.4936, 0.3147, 0.2843, 0.2416, 0.2127, and 0.20. We used five (5) different classification algorithms (J48_Consolidated, LMT, RepTree, MultiBoost Decision Stump, and Random Forest) and their results were compared. Performance analysis on the five (5) classifiers shows that Multiboost Decision Stump exhibited the best accuracy of 94% and least MAE of 0.0604. The findings show that the developed model when integrated into automated disease diagnostic workflows will provide an enhanced and flexible solution for quick prognosis of typhoid fever in children in the tropics thereby reducing the cost of hospitalization among others.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reducing the Financial Burden of Typhoid Fever Treatment Through Intelligent Modeling

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Tosin C. Olayinka,
  • Akinola S. Olayinka,
  • Ojei H. Onyijen,
  • Edwin Onatuyeh,
  • Wilson Nwankwo,
  • Pascal C. Nwankwo

摘要

Typhoid fever remains a significant public health challenge in Africa particularly in regions with limited access to clean water and sanitation. The financial burden associated with typhoid fever treatment can exacerbate disparities in healthcare access and affordability, especially in resource-constrained settings. By leveraging data analytics and machine learning techniques, this paper seeks to develop predictive models capable of accurately assessing the risk of typhoid fever and guiding resource allocation strategies. Through the integration of diverse datasets, including clinical parameters, demographic information, and environmental factors, the proposed modeling framework aims to identify key predictors of typhoid fever incidence and severity. The dataset comprises laboratory tests of blood samples of one hundred and fifty (150) children, from age zero (0) to five (5) years. Data preprocessing and training were performed using Waikato Environment for Knowledge Analysis (WEKA) workbench tool as a machine learning platform. The information Gain (IG) feature selection technique was employed to select the best attributes using a cut-off gain of 0.2. The top six (6) attributes with more information gain relevant to the class typhoid were selected. Widal test, Malaria count, Monocyte, Platelets, HB, Eosinophils, cost of treatment are the selected attributes with their respective ranked information gain of 0.6759, 0.4936, 0.3147, 0.2843, 0.2416, 0.2127, and 0.20. We used five (5) different classification algorithms (J48_Consolidated, LMT, RepTree, MultiBoost Decision Stump, and Random Forest) and their results were compared. Performance analysis on the five (5) classifiers shows that Multiboost Decision Stump exhibited the best accuracy of 94% and least MAE of 0.0604. The findings show that the developed model when integrated into automated disease diagnostic workflows will provide an enhanced and flexible solution for quick prognosis of typhoid fever in children in the tropics thereby reducing the cost of hospitalization among others.