Machine learning model for predicting malaria using clinical information.

Area of Science:

• Medical Informatics
• Computational Biology
• Epidemiology

Background:

• Rapid malaria diagnosis is critical for disease control.
• Traditional blood smear image analysis for malaria diagnosis has limitations.
• This study explores alternative diagnostic methods using patient information.

Purpose of the Study:

• To develop and evaluate machine learning models for malaria diagnosis using patient data.
• To compare the performance of various machine learning algorithms.
• To identify key patient information features for malaria prediction.

Main Methods:

• Extracted patient information from PubMed abstracts (1956-2019) to create datasets.
• Compared six machine learning models: Support Vector Machine, Random Forest (RF), Multilayered Perceptron, AdaBoost, Gradient Boosting (GB), and CatBoost.
• Utilized Synthetic Minority Oversampling Technique (SMOTE) to address data imbalance.

Main Results:

• For parasitic disease-only datasets, RF was the top-performing model, even without SMOTE.
• For the total dataset, GB initially performed best, but RF excelled after SMOTE application.
• Nationality emerged as the most crucial feature for malaria prediction in imbalanced data, while symptoms were key in balanced data with SMOTE.

Conclusions:

• Machine learning models can be successfully applied for malaria prediction using patient information.
• The choice of the best model and important features depends on dataset composition and balancing techniques.
• This approach offers a promising alternative to traditional malaria diagnostic methods.