Noise-aware celestial clustering for hot topic detection from microblog datasets with not well-separated topics

Due to the large volume of data, presence of high noise and abundance of similar topics, hot topic detection algorithms find it difficult to detect topics accurately from microblog datasets. Here, we propose a novel clustering technique that is able to group hot topics from microblogs even in the presence of large amount of noise and similar topics. We use a combination of centroid- based celestial clustering (CBCC), a nature-inspired clustering mechanism derived from the principles of particle physics, and PSO- based approach to achieve this. We experiment our methods on four different microblog datasets, two of which include a large number of outliers. We have evaluated our methods using statistical parameters such as precision, recall, error rate, \(F_1\) $F_1$ score and average in group proportion (AIGP) and comparison with other popular topic detection algorithms has shown that our methods yield significant advantages.