Feature based opinion analysis on social media tweets with association rule mining and multi-objective evolutionary algorithms

Social media platform has achieved wide popularity in presenting the user-generated information online. The proliferation of user-generated content through social networking sites can enhance the existing transportation system. This work adds to the existing research by proposing a novel system to assess transit rider's reviews on the quality of transport services using Twitter information. A novel framework of a multi-objective evolutionary approach with association rule mining is proposed for feature-based opinion analysis on social media transit reviews. Feature-based opinion analysis is performed in two steps such as feature extraction and opinion analysis. First, the corpus of association rules is generated with opinion analysis, morphological and syntactic analysis, and semantic representation. Then, multi-objective flower pollination, multi-objective gray wolf, multi-objective moth flame, and multi-objective cat swarm optimization are applied to discover high-quality association rules on the transit user's opinion. The experimental results indicate that multi-objective cat swarm optimization using association rule mining performs better in terms of confidence, coverage, the interestingness of the rules, computational time, mean average support, and mean average confidence than the other proposed as existing approaches. It is observed that MCSO-ARM achieved an improvement in the range of 0.21%-0.27% in terms of confidence, 0.10%-0.13% in coverage, 0.10%-0.25% in interestingness, the computational time of 27.5 s, mean average support value of 0.10%-0.30% and mean of 0.20%-0.30% compared to the existing MPSO-ARM and other proposed MGWO-ARM, MMFO-ARM, and MFPO-ARM approaches.