Brand Potential User Identification Algorithm Based on Sentiment Analysis

Area of Science:

• Natural Language Processing
• Machine Learning
• Data Mining

Background:

• Sentiment analysis and potential customer identification are crucial for businesses.
• Existing methods for sentiment dictionary construction have limitations.
• Social media data presents challenges due to its volume and imbalance.

Purpose of the Study:

• To develop an improved method for constructing domain-specific sentiment dictionaries.
• To propose a joint sentiment topic model for potential customer identification.
• To address data imbalance issues in prospect identification tasks.

Main Methods:

• Developed an improved integration rule for sentiment dictionary construction.
• Proposed an automatic construction method for domain sentiment dictionaries in social media.
• Integrated domain sentiment lexicons into a joint sentiment topic model.
• Implemented sample resampling and a diverse integration framework for imbalanced data.

Main Results:

• The proposed domain sentiment lexicon construction method demonstrated good performance.
• The joint domain sentiment topic-based lead identification method showed effectiveness.
• The methods successfully addressed data skewing and category imbalance.
• Experimental validation on a social media corpus confirmed the approach's efficacy.

Conclusions:

• The study provides a robust framework for domain sentiment lexicon construction.
• The joint sentiment topic model offers a valuable solution for prospect identification.
• The research holds both theoretical significance and practical value for businesses.