A Computational Intelligence based Early Diagnosis of Asthma Disease: A Saudi Arabian Case Study

Abstract

<title>Abstract</title> According to recent research, an increase in chronic diseases including Asthma has been identified. The number of Asthma patients in the Kingdom of Saudi Arabia is a cause for concern due to the weather conditions and lifestyle especially in post-pandemic era. This led to the consideration of a solution to decrease the number of infections, and one of the solutions is developing an active system to detect Asthma at its early stage, thereby preventing the disease or providing early treatment to cure it. Inthe current study, we have utilized machine learning to develop tools that serve as early warning signs to track the subtlest symptoms of Asthma at a very early stage (pre-symptom stage). We also aimed for the highest achievable precision using the least number of features. Although there have been various prior attempts to apply machine learning to predict the occurrence of Asthma. Nevertheless, focusing on the identification of the disease at the (pre-symptom stage) and specifically with the Saudi Arabian data has never been reported based on our investigation of the reviewed literature. The dataset for the current study was obtained from King Fahad University Hospital, Dammam, Saudi Arabia according to standard tests among the patients including blood tests, virus tests, and biochemistry tests. The dataset contains seventeen significant attributes and includes information for 328 Asthma patients with 165positives and 163negatives. The methods selected for application here are Random Forest (RF), Artificial Neural Network (ANN), Support Vector Machine (SVM), and Naive Bayes (NB). Each of these methods has been chosen based on its distinctive features. The experimental outcome revealed that the RF, SVM, and ANN approaches yielded to 94%, which is the highest accuracy and improved compared to state-of-the-art by 3.9%. It is worth noting that nine features out of seventeen possible features were used to create the above accuracy. Despite RF, SVM, and ANN approaches generating the same accuracy, ANN has a higher cost of error. Therefore, RF and SVM are superior based on the pattern of results achieved, and hence, they are suggested for this particular problem.