A Framework Questionnaire for Diagnosing Infectious Disease Using Machine Learning Techniques

Infectious diseases weigh down the communities in the world and scientists to spend more effort via keeping tracking of evolving treatment and detecting methods. These diseases may lead to harm life of people. Early diagnosis could significantly support healthcare specialists to save more lives. Additionally, the pandemic leads to maximizing hospitals visits, medical clinics and healthcare centres. The international health organizations also have shown that there has been a rapid growth of infected cases. Therefore, correct diagnosis has become a pressing problem. Consequently, automated diagnosis becomes a sensible solution to the problem of these diagnosis challenges. This study was conducted to identify the most common infectious diseases in the Iraqi society using a well-designed questionnaire and a proposed automated diagnostic technique. Firstly, the top diseases questionnaire is distributed around the city of Baghdad to different medical clinics. The results from the preliminary analysis of the collected responses (115 responses) showed that the most common widespread diseases in the Iraqi community are diabetes, flu, and typhoid. This was followed by another questionnaire for the identification of symptoms and blood test variables for these diseases. It is worth pointing out that there are not sufficient and updated studies dealing with the diseases that attack the Iraqi community. Toward the automated diagnosis, both infectious diseases (flu and typhoid), identified symptoms are employed as feature space with one of the machine learning techniques. For the results evaluation different measures, such as accuracy, confusion matrix, and efficient verification via ROC, have been used to indicate the system performance. The result shows that typhoid disease has significant diagnosis accuracy of 98% compared to the others. While three machine learning systems named (Native Bayes, Linear discriminant, and Ensemble (subspace discriminant)) were used to diagnose flu disease. The resulting accuracy of all three models are 92% which shows good performing. Therefore, the proposed method shows precise accuracy and systematic manner for analyzing infectious diseases.