TY - JOUR T1 - Proposing an effective technological solution for the early diagnosis of COVID-19: a data-driven machine learning study TT - پیشنهاد یک راه‌کار فناورانه موثر جهت تشخیص زودهنگام بیماری کووید-19: مطالعه مبتنی بر یادگیری ماشین داده محور JF - hums-jmis JO - hums-jmis VL - 7 IS - 1 UR - http://jmis.hums.ac.ir/article-1-284-en.html Y1 - 2021 SP - 68 EP - 78 KW - COVID-19 KW - Coronavirus KW - machine learning KW - data mining KW - diagnostic model. N2 - Aim: Accurate and timely diagnosis of COVID-19 using artificial intelligence and machine learning technologies will play an important role in improving the disease indicators, optimal utilization of limited hospital resources and reducing the burden on pandemic healthcare providers. Therefore, this study aimed to evaluate the efficiency of selected data mining algorithms based on their performance for COVID-19 diagnosis. Methods: The present study was a retrospective applied-descriptive study that was conducted in 2020. In this study, the data of patients admitted with a definitive diagnosis of Covid-19 from March 17, 2020 to December 10, 2020 were extracted from the Electronic Medical Record (EMR) database in Ayatollah Taleghani Hospital in Abadan. After applying the inclusion and exclusion criteria to identify the samples, 400 records were entered into the data mining software. The data were compared using chi-square criterion to determine the variables of teach algorithms and their performance based on different evaluation criteria in the turbulence matrix. Results: Comparing the performance from data mining algorithms based on different evaluation criteria in the turbulence matrix revealed that the J-48 algorithm with the sensitivity, precision, and Matthews Correlation Coefficient (MCC) of 0.85, 0.85 and 0.68 respectively had better performance than the other data mining algorithms for the disease diagnosis. The 3 variables of lung lesion existence, fever, and history of contact with suspected COVID-19 patients, by considering Gini Index to determine the point of division, with Gini index of 0.217, 0.205 and 0.188 respectively were considered as the most important diagnostic indicators of COVID-19. Conclusion: Using selected data mining methods, particularly J-48 algorithm will greatly aid the timely and effective diagnosis of COVID-19 in the form of clinical decision support systems. M3 10.52547/jmis.7.1.68 ER -