Scholarlink

<p>To evaluate the utility of Clinical signs and symptoms utilising machine-learning algorithms in detecting malaria. Data on patient symptoms that are relevant to the diagnosis of malaria were gathered from real-time clinical records from a medical college hospital. Preprocessing techniques were used to guarantee the consistency and quality of the data. This comprises scaling numerical features, coding categorical variables, and missing value management as required. A labelled dataset with patient symptoms and associated malaria diagnosis outcomes was trained on the chosen machine-learning models. The testing data was used to assess each model’s performance using standard classification measures, such as accuracy, recall, and AUC score after it had been trained using the training set. For this study, eight supervised machine learning models, namely Extra Trees Classifier, AdaBoost, XGBoost, Random Forest, Logistic Regression, Support Vector Machine, Naive Bayes, and Artificial Neural Network, were evaluated using both single-split and tenfold cross-validation. The Extra Trees Classifier emerged as one of the strongest models for symptom-based malaria prediction, achieving the highest mean AUC (0.9594 ± 0.0255) along with balanced accuracy (0.9038 ± 0.0412) and recall (0.8240 ± 0.0709), indicating robust and reliable performance across folds. This paper proposes a region-specific malaria prediction model for clinical diagnosis that uses patient demographics and clinical symptoms. This model has the potential to assist in the development of a clinically based malaria diagnostic system.</p>

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

Leveraging machine learning for symptom-based malaria diagnosis: a predictive model approach

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • S. Ruban,
  • Varenya Vinay,
  • Hilda Shanthini,
  • Sanjeev Rai,
  • SivaKannan Subramani

摘要

To evaluate the utility of Clinical signs and symptoms utilising machine-learning algorithms in detecting malaria. Data on patient symptoms that are relevant to the diagnosis of malaria were gathered from real-time clinical records from a medical college hospital. Preprocessing techniques were used to guarantee the consistency and quality of the data. This comprises scaling numerical features, coding categorical variables, and missing value management as required. A labelled dataset with patient symptoms and associated malaria diagnosis outcomes was trained on the chosen machine-learning models. The testing data was used to assess each model’s performance using standard classification measures, such as accuracy, recall, and AUC score after it had been trained using the training set. For this study, eight supervised machine learning models, namely Extra Trees Classifier, AdaBoost, XGBoost, Random Forest, Logistic Regression, Support Vector Machine, Naive Bayes, and Artificial Neural Network, were evaluated using both single-split and tenfold cross-validation. The Extra Trees Classifier emerged as one of the strongest models for symptom-based malaria prediction, achieving the highest mean AUC (0.9594 ± 0.0255) along with balanced accuracy (0.9038 ± 0.0412) and recall (0.8240 ± 0.0709), indicating robust and reliable performance across folds. This paper proposes a region-specific malaria prediction model for clinical diagnosis that uses patient demographics and clinical symptoms. This model has the potential to assist in the development of a clinically based malaria diagnostic system.