Emotion analysis of social network data using cluster based probabilistic neural network with data parallelism

Social media contains a huge amount of data that is used by various organizations to study people's emotions, thoughts and opinions. Users often use emoticons and emojis in addition to words to express their opinions on a topic. Emotion identification from text is no exception, but research in this area is still in its infancy. There are not many emotion annotated corpora available today. The complexity of the annotation task and the resulting inconsistent human comments are a challenge in developing emotion annotated corpora. Numerous studies have been carried out to solve these problems. The proposed methods were unable to perform emotion classification in a simple and cost-effective manner. To solve these problems, an efficient classification of emotions in recordings based on clustering is proposed. A dataset of social media posts is pre-processed to remove unwanted elements and then clustered. Semantic and emotional features are selected to improve classification efficiency. To reduce computation time and increase the efficiency of the system for predicting the probability of emotions, the concept of data parallelism in the classifier is proposed. The proposed model is tested using MATLAB software. The proposed model achieves 92 % accuracy on the annotated dataset and 94 % accuracy on the WASSA-2017 dataset. Performance comparison with other existing methods, such as Parallel K-Nearest Neighboring and Parallel Naive Byes Model methods, is performed. The comparison results showed that the proposed model is most effective in predicting emotions compared to existing models.