부산대학교 도서관

The proliferation of social media platforms has ushered in a deluge of user-generated content, encapsulating vast sentiments and trends that shape public discourse. This research endeavors to harness these digital traces through sophisticated data mining techniques and predictive analytics to distill sentiments and forecast trends from social media datasets. Leveraging state-of-the-art Natural Language Processing (NLP) algorithms, the study develops a robust framework that systematically identifies, extracts, and analyzes affective states and opinions embedded within textual data. The novel sentiment analysis model proposed here demonstrates significant advancements over traditional lexicon-based and machine learning approaches by incorporating contextual embeddings and deep learning architectures, enhancing the granularity and accuracy of sentiment classification. Furthermore, the paper presents an innovative trend prediction methodology that combines time-series analysis with social network theory to predict emergent topics and shifts in public opinion. This predictive model is validated through extensive experiments on diverse social media platforms, showcasing its efficacy in real-time scenario simulations. The implications of this work are manifold, providing valuable insights for businesses, policymakers, and researchers in understanding the zeitgeist and its dynamics. The research not only contributes to the academic discourse on sentiment analysis and trend prediction but also serves as a bellwether for practical applications in market analysis and sociopolitical strategizing.