Extracting Hidden Patterns of Iranian User Trust in Social Networks Regarding Coronavirus Disease 2019 Using Data Mining Techniques

Abstract

Abstract Aim: The coronavirus disease 2019 (COVID-19) pandemic caused the use of social networks in the field of information acquisition and transmission to increase, whereas the validity of the information available is questionable. Because people’s trust in these networks is important, this study aimed to utilize three data mining techniques to identify the hidden rules for detecting the user trust level of social networks in the context of COVID-19. Materials and Methods: An electronic questionnaire containing 27 questions was provided to users. Out of the 12 questions selected, the final question asked about the level of user trust in social networks and was considered the target class. Based on the range in value, question 12 was divided into five classes. The relevance of the remaining 11 questions was then assessed using three decision tree-based data mining techniques. Results: The results showed that the random forest technique performed better than the other techniques. Most social network users have a moderate level of trust in information regarding COVID-19; in fact, the medium class is the most widely used target class with 60% utilization rate, which affects sensitivity and specificity. The values of these measures were much higher for this class than for the other classes. Conclusion: The educational content, both its type and the amount, regarding COVID-19 that is provided on social networks, impact on user trust. As the existence of inconsistent information has had a negative impact on user trust, a small percentage of users have high trust in these networks.