An Investigative Process Model for Predicting Information Difusion on Social Media: Information System Perspective
The challenge of the information diffusion (ID) process across social media is the inability to know the provenance or source of information and the details of the person sending the information. This situation creates confusion within the organization of people, which hinders smooth communication. In this paper, a classification and a mathematical model for predicting ID on social media (SM) were formulated. This was with a view to know the source of tweeted information for clarity and decision-making. The formulated model was simulated and evaluated to ascertain its adequacy. Data were collected from Twitter by using the Twitter Search API. In particular, 3,200 tweets were extracted on related political matters in Nigeria. The results showed that the classification is a multiclass case. The macro-average ROC curve from it showed 0.93. The ROC from curve one to seven of each class is 0.90, 0.95, 0.98, 0.97, 0.90, 0.97 and 0.85, respectively. The results of the simulation also indicated that there were seven different types of source devices for the information, namely Android, BlackBerry, iPad, iPhone, rack, studio, and web, according to the dataset. Most of the information came from iPhones, with about one thousand, nine hundred and eighty-four (1,984) tweets out of the 3,200 (that is 62%); the studio was the source least frequently with seven tweets (that is, 0.22%). In terms of the rate of diffusion, information gets most diffused through iPhones (76.20% of the retweets), and least diffused with Android phones (0.56% of the retweet). The model evaluation shows an accuracy of 95.87%, a specificity of 97.64% and a sensitivity of 83.46%, making it suitable for prediction. The research not only unveiled the identity of devices used for sending information on SM but also revealed the rate at which information gets diffused with the devices to aid decision-making.