UBIQUITOUS LEARNING-DRIVEN NATURAL LANGUAGE PROCESSING FOR TWITTER CLIMATE CHANGE SENTIMENT ANALYSIS USING BAYESIAN MACHINE LEARNING FOR FOSTER SUSTAINABLE DEVELOPMENT IN RURAL AREAS

Abstract

The size of data shared has promptly surged with the usage of social networks, and this has become a crucial research area for environmental concerns. Sentiment analysis (SA) defines the behavior and people’s sensitivity to environmental problems. Ubiquitous learning and natural language processing (NLP) intersect to address climate change SA by analyzing vast textual data from scientific reports, social media, and news to gauge public attitudes and perceptions. This provides meaningful insight for organizations and policymakers working on climate change mitigation systems in rural areas and allows local communities to engage in environmental discourse actively, fostering a bottom-up algorithm to address climate problems. This study develops a new twitter climate change sentiment analysis using the Bayesian machine learning (TCCSA-BML) technique to promote sustainable development in rural areas. This technique exploits ubiquitous learning with NLP technologies to identify climate change in rural areas. Also, the TCCSA-BML technique undergoes data preprocessing in several ways to make the input data compatible with processing. Besides, the TCCSA-BML technique utilizes the TF-IDF model for the word embedding process. Moreover, the classification of various kinds of sentiments occurs using the Bayesian model averaging (BMA) technique comprising three classifiers, namely attention long short-term memory (ALSTM), extreme learning machine (ELM), and gated recurrent unit (GRU). Finally, the parameter tuning of the classifier is implemented by the coyote optimization algorithm (COA) model. The performance analysis of the TCCSA-BML approach is evaluated on the Kaggle SA dataset. The experimental validation of the TCCSA-BML approach portrayed a superior accuracy value of 94.07% over other models.