Unsupervised sentiment analysis is of great significance to researchers who rely on automated artificial intelligence tools to learn from unlabeled reviews on social media where people express their opinions on various issues or events. Lack of ground truth annotations renders supervised learning to be unsuitable for sentiment analysis. Some popular unsupervised sentiment analysis tools are AFINN, VADER, and TextBlob. In the current work, we propose an ensemble of these three techniques by performing fuzzy aggregation of the scores generated. The sentiment polarity scores predicted by the three methods are subject to the fuzzy aggregation procedure, and the resulting aggregated score is used to predict the sentiment. Experiments on the 50K samples of the benchmark IMDB Movie reviews dataset containing an even number of positive and negative reviews reveal that fuzzy aggregation improves the scores of the individual models, thereby motivating the future development of hybrid fuzzy models for the sentiment analysis task.