فهرست نشریات علمی دسترسی آزاد ایران

Backgrund and Objective: The COVID-19 pandemic, an infectious and airborne disease, has been classified as an acute respiratory syndrome by the World Health Organization (WHO). Computer technology in medicine has emerged as an appropriate method for early diagnosis of this disease. Timely and accurate diagnosis of COVID-19 using machine learning algorithms plays a vital role in disease management. Despite the availability of much data on the epidemic status of this disease from various sources, and the utilization of artificial intelligence for diagnosis and prevention in many studies, obtaining accurate and reliable data remains a challenge. Therefore, the aim of this study is to predict, evaluate, compare, and analyze the performance of Jaccard, Pearson, and Cosine similarity criteria algorithms using real hospital datasets while addressing the challenge of obtaining accurate and real data.Methods: This study was conducted using outpatient and inpatient data from Mousabn-e-Jafar Charity Hospital in Mashhad, Iran, spanning from May 2020 to July 2022, to investigate the application of machine learning systems and similarity criteria for intelligent diagnosis of COVID-19. A dataset[1] was compiled and uploaded to online repositories, and four methods of COVID-19 diagnosis were utilized. The symptoms of hospitalized patients were compared with the COVID-19 dataset, and a similarity matrix was created to calculate the similarity of all patients. Classical similarity techniques were then applied to the data, and finally, the most effective algorithm was determined based on the comparison of evaluation criteria.Results: The Jaccard similarity method demonstrated the highest value of 0.94 in the recall criterion, while the accuracy criterion showed 1 in the Pearson similarity method and 0.76 in the Jaccard similarity method, with an F1 score of 0.86. The crucial factors for diagnosing the disease were found to be white blood cells (WBC), platelets, RT PCR, CT SCAN, shortness of breath, fever, SPO2, and the number of breaths per minute. These factors were identified as essential in the diagnosis of COVID-19 based on the analysis using the Jaccard and Pearson similarity methods.Conclusion: Based on the results obtained from this study, the real dataset extracted has shown great potential in assisting with the diagnosis and prediction of COVID-19. This dataset may be valuable for researchers in computer science, medicine, epidemiology, and other applied fields. The findings of this study could contribute to further research and advancements in the field of COVID-19 diagnosis and prediction, and could be beneficial in various perspectives and applications beyond medicine.