Twitter is one of the most popular social networks with millions of audience every day around the world. Analyzing and predicting the popularity of Twitter posts have been one of widely studied topics that allows us to uncover the pattern and trend of collective interest of Twitter audience towards each post. In this study, we present a novel method in analyzing and predicting temporal profile of Twitter popularity, including both retweets and replies, using dynamic time warping (DTW) in finding the similarity among the temporal profiles of the popularity. Then, similar temporal profiles are grouped together using sequential clustering and the centroids of each cluster are determined by calculating Barycenter of each cluster. These centroids are used as popularity profile templates for the popularity prediction of a new post. The proposed method is tested with real Twitter posts obtained from international Twitter news channels. The experimental results show that the proposed method outperforms the existing method which is based on exponential model.