Microblogs are growing at an overwhelming rate due to the liberty of content generation. This led to a diversity of misleading use, such as digitally manipulated content and reposting of real content in a wrong context. In order to remedy such deficiencies numerous solutions were proposed in the field of credibility evaluation of information shared on microblogs. In this paper, we review previous works about tweet credibility assessment and introduce a new approach that allows for the measurement of tweet credibility using Random Forest Regressor. In our approach, we consider credibility as a continuous value instead of a binary one and rely on three features namely; accordance between tweet and event sentiment, tweet grammar, presence of URL to predict credibility. These features were considered by the literature as suitable for credibility prediction. Based only on content features, our approach is able to predict credibility values with an out of bag score of 86.46 %.