OPTIMIZING SOCIAL MEDIA SENTIMENT ANALYSIS USING HYBRID DATA MINING TECHNIQUES FOR ENHANCED MARKETING INTELLIGENCE

Abstract

The rapid growth of social networking platforms has created a vast digital landscape rich with consumer opinions, preferences, and behaviors. This thesis presents a hybrid approach to social media sentiment analysis, aiming to enhance marketing intelligence by integrating textual, contextual, and behavioral data mining techniques. Traditional sentiment classification methods, often based solely on text, fail to account for the dynamic nature of social platforms. In response, this research proposes a model that combines rule-based ensemble sentiment scoring with supervised machine learning and probabilistic models, such as conditional random fields (CRFs) and Monte Carlo Markov Chains (MCMC). The framework also incorporates distributed computing to handle large-scale datasets, ensuring scalability and efficiency. Experimental validation using Facebook and Twitter datasets demonstrates improved accuracy in identifying sentiment polarity, ranking social entities, and targeting specific audience segments. The results affirm that integrating syntactic and semantic cues with user behavioral data offers a deeper understanding of online discourse. This approach enables marketers to deploy more precise and adaptive strategies, thus transforming raw social data into actionable business intelligence. The work contributes to the growing field of big social data analytics by offering a scalable, intelligent, and contextaware sentiment analysis framework for targeted digital marketing