Sentiment analysis in social network data using multilayer perceptron neural network with hill-climbing meta-heuristic optimisation

Social networks such as Twitter, Facebook, and Instagram are the most widely used communication media in which users share their feelings and opinions about current events in society, for example, the occurrence of COVID19, its causes, symptoms, precautions, safety measures, prohibitions, etc. This work proposes a multi-class sentiment classification model to classify the tweets under various polarisation categories of the social network data. For Twitter data classification, this work proposes a model based on a multilayer perceptron neural network with hill-climbing optimisation. The heuristic hill climbing is used at the backpropagation for learning. TF-ID method is used for feature extraction from the dataset. This work compares the proposed method with the existing sentiment classification models of social network data. The proposed multi-class sentiment classification model has shown improvements in performance with the measures namely f1-score, precision, accuracy and recall over the existing sentiment classification models.