Enhancing User Sentiment Analysis of Social Media Reviews using Fuzzy Inference Opinion Mining and Deep Learning for Predicting Consumer Reconstruction Intent

Social media's explosive growth has made it an invaluable tool for user Sentiment Analysis (SA) and buyer behavior research. However, several obstacles must be overcome to extract valuable data from the vast amount of unstructured social media reviews, especially when accurately predicting customer reconstruction intentions. This paper presents a unique solution to these challenges by combining fuzzy reasoning with a hybrid Deep Convolutional Neural Network-Multi BiDirectional Long Short-Term Memory (DCNN-mBiLSTM) architecture to enhance customer SA through opinion mining. The mBiLSTM captures long-term dependencies from text reviews, enabling more precise intent forecasting, while the DCNN extracts local features from the reviews. Fuzzy logic is incorporated to address uncertainty and unpredictability in customer attitudes, improving the accuracy of sentiment classification. The study aims to enhance SA and intent forecasting precision while providing valuable insights into users' reconstruction intentions. The results demonstrate that the proposed DCNN-mBiLSTM algorithm, combined with fuzzy inference, significantly outperforms Existing SA techniques in predicting preferences and forecasting customer behavior.