A Comparative Study of Text Classifier for Mobile Crowdsensing Applications

摘要

Mobile reporting applications are useful mainly for reporting real-time issues related to public infrastructure, environmental or social incidents through smart mobile devices. The credibility of the cases reported are often a great challenge because users may report false informationand as a result this affects the response team in the aspect of time, energy and other resources. Researchers in the past have developed many report trust estimation algorithms that focuses on user’s location, behavior and reputation. We aim to analyze the textual part of a report. Textanalyses have been used for email spam filtering and sentiment analysis but have not been used for false report identification. Therefore, the purpose of this study is to compare different text classification algorithms and propose a suitable classifier for distinguishing the genuine and fakereports. The comparative analysis can be used by other researchers in the area of false report or fake message identification.