Cancer Research Trend Analysis Based on Fusion Feature Representation

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Machine Learning

Background:

• Effective text feature representation is crucial for machine learning in biomedical research.
• Current methods often rely on word representation, which may not fully capture semantic and structural information.
• Discovering research trends aids knowledge dissemination and guides future studies.

Purpose of the Study:

• To propose novel fusion algorithms for enhanced text feature representation.
• To evaluate the effectiveness of these algorithms compared to classical models.
• To apply the best-performing algorithm for trend analysis in cancer research.

Main Methods:

• Development of two fusion algorithms: Tr-W2v and Ti-W2v, building upon classical text representation models.
• Incorporation of word importance into the text feature representation process.
• Application of the Tr-W2v algorithm for correlation analysis, keyword trend analysis, and improved keyword trend analysis in cancer research.

Main Results:

• The proposed Tr-W2v and Ti-W2v algorithms significantly outperform classical text representation models.
• The Tr-W2v algorithm achieved the best results in text feature representation.
• Trend analyses revealed insights into cancer research hotspots and evolution.

Conclusions:

• The Tr-W2v and Ti-W2v algorithms offer superior text feature representation for biomedical research.
• The Tr-W2v algorithm is effective for identifying research trends and evolution in specific domains like cancer research.
• These findings can assist researchers in navigating and directing future scientific endeavors.