Persian Rumor Detection Using a MultiClassifier Fusion Approach

During the last few years, rumor and its rapid diffusion via social media have affected public opinions, even in some important such as presidential elections. One of the main approaches for rumor detection methods is based on content and natural language processing. Despite considerable improvement made in this regard in the English language, unfortunately, we have not witnessed enough progress in the Persian language, mainly due to a lack of datasets in this area. The main novelty of this paper is combining different learning methods to consider the classification problem from different aspects and combine the classifiers’ results to achieve a reasonable final result. In the proposed method, each classifier is assigned a weight depending on its f-measure value; thus, the final fused result is closer to the performance of the best classifier. When news samples have various characteristics, and the best classifier is not predetermined, this fusion method is more beneficial. Therefore, as the conclusion of this research, compared to a single rumor detection method, the fusion of classifiers could be used to achieve better results when the news samples have various characteristics.

کلیدواژه ها:

Rumor detection ، Machine learning ، Content-based text classification ، Deep learning ، Multi-classification