Hybrid model of machine and deep learning to analyze Twitter data and prediction of online public opinion: revisiting agenda-setting implications

Purpose This paper aims to examine the impact of social media on online public opinion – an area with limited existing studies. The study is undertaken to bridge the gap. It introduces the hybrid model for predicting online public opinion about political parties in Pakistan. The model analyzes user emotions and behavior, encompassing anger, fear, joy, anticipation, healthcare and education policies. Design/methodology/approach This study employs a communication framework to examine the interplay between social media dynamics and Pakistan’s political landscape. The proposed hybrid model combines supervised machine learning and deep learning for Twitter-based text classification to assess political perceptions. A total of 52,067 tweets were analyzed and categorized into eight emotion groups. Findings The study reveals a significant increase in public engagement with social and political discussions on social media. Politicians increasingly leverage social media platforms as tools for communication and propaganda, influencing democratic norms. The proposed hybrid model effectively classified political sentiments and demonstrated high accuracy in analyzing public perceptions of political parties in Pakistan. Practical implications This research proposes a cost-effective communication framework to enhance public engagement. Political communication offers a methodological baseline for policymakers to understand public opinion and design their communication strategies using prediction models. Originality/value The study introduces a hybrid approach to evaluate political sentiments by analyzing user emotions and behavior. The comparison of healthcare and education policies adds depth to the analysis and provides valuable information for voters. This research extends agenda-setting theory in the digital era by providing robust and accurate forecasting of policymakers’ agendas which were previously limited. Using machine and deep learning, this research reveals that agenda setting is a reciprocal and dynamic process and is shaped by the digital interplay between media, policymakers and public response.