ANALYZING CONSUMER PREFERENCE DYNAMICS IN AMAZON REVIEW DATA USING DOUBLE PROPAGATION-BASED SENTIMENT ANALYSIS

This research proposes an innovative approach to extracting consumer preferences using Sentiment Analysis, addressing the limitations of traditional methods like Conjoint Analysis. Consumer preferences, influenced by factors such as product quality, experience, and emotions, are crucial for shaping marketing strategies and product development. While Conjoint Analysis faces issues like small sample sizes and high costs, Sentiment Analysis, particularly Aspect-Level Sentiment Analysis (ABSA), provides a more detailed and dynamic method by analyzing sentiments toward specific product attributes. Leveraging big data from e-commerce reviews, this technique offers real-time insights, is cost-effective, and does not require predefined sentiment dictionaries. The study utilizes the Double Propagation (DP) technique, which enhances sentiment accuracy through bidirectional analysis. This approach offers a more efficient, automated, and adaptive solution for understanding consumer preferences. The evaluation, conducted on Amazon Review Data, confirms the moderate effectiveness of DP in aspect extraction, with room for enhancement in dependency rules and aspect identification algorithms. The results of experiment also highlight the variability in DP's aspect-level accuracy, with consistently strong results for Price and Performance, but challenges in Port, Design, and Storage.