A performant deep learning model for sentiment analysis of climate change

Climate change is one of the most trend topics of the decade in the world. The recent years were the warmest in 139 years, however identifying deniers and believers of this subject still a very big issue. The challenge is to have an efficient tool to detect deniers in order to deploy the appropriate strategy facing this phenomenon. Moreover, Bidirectional Encoder Representations from Transformers (BERT) pre-trained model has taken Natural Language Processing tasks results so far. In this paper we presented an efficient technological tool based on deep learning model and BERT model for detecting people's opinions on climate change on social media platforms. We used convolutional neural network targeting the public opinions on climate change on Twitter. The results showed that our model outperforms the machine learning approaches: Naive Bays, Support Vector Machine and Logistic Regression. This model is able to analyze people's behavior and detect believers and deniers of this disaster with high accuracy results (98% for believers and 90% for deniers). Our model could be a powerful citizen sensing tool that can be used by governments for monitoring and governance, especially for smart cities.