Sampling Technique to Reduce Execution Time Using Multi-Viewpoints Cosine-Based Similarity Metrics with Hybrid Topic Models in Social Network Datasets

Twitter is the most widely used social networking platform to extract high-quality topics from a Twitter corpus-based on text pre-processing quality and the detection of an appropriate cluster tendency of different types of datasets. In Twitter-based tweets, fewer frequent words were present, because of which identification of cluster patterns becomes more challenging by using traditional topic models. Traditional topic models used Euclidean distance metrics in most of the techniques and cosine-based distance metrics with a single viewpoint in others to manually estimate the k values. In this work sampling technique is used to reduce execution time in finding the k values of a particular dataset with multi-viewpoints cosine-based similarity metrics with hybrid topic models. This visual technique first extracts the similarity characteristics of data objects by using multi-viewpoints cosine-based similarity metric and also with cosine and Euclidean distance metrics and then a sampling technique is applied to reduce execution time. The proposed Sampling Multi-Viewpoints Visual Method Visual Assessment Tendency (SMVP-VM-VAT) method results were compared with traditional topic models and proved the proposed method performed well than that of other models.