Trending topic prediction by optimizing K-nearest neighbor algorithm

Quick and accurate decision-making has currently been a modern governmental characteristic, since the development of information and communication technologies also grow very rapidly. One way to support the quick and accurate decision-making is predicting the trending topic. The research aimed at assisting the government of Makassar City to predict the trending topic which would happen by analyzing the historical stack in the data mining. The method used was K-Nearest Neighbor (KNN), in which prediction on the trending topic was determined based on the membership distance of a class. The research was conducted based on the news and conversation taken from the online and social media related to Makassar City Government with 393.667 raw data, in which the preprocessing was then carried out to determine the trending and non-trending conversations, producing 2007 trained and tested data. The system performance analysis technique applied was confusion matrix with calculation of percentages of the accuracy, precision, and recall. The result showed that using K-Nearest Neighbor (KNN), the accuracy of 81,13% is obtained.