Symbolic representation based on trend features for biomedical data classification.

Area of Science:

• Biomedical data analysis
• Machine learning in healthcare
• Data mining techniques

Background:

• Increasing volume of biomedical data from portable devices exceeds human analytical capacity.
• Computer-assisted analysis is crucial for modern medical diagnosis.
• Symbolic representation offers advantages in noise elimination and dimensionality reduction for data mining.

Purpose of the Study:

• To improve biomedical data classification using symbolic representation.
• To leverage the benefits of discrete data and dimensionality reduction for enhanced data mining.
• To introduce a novel method for symbolic representation of biomedical data.

Main Methods:

• Introduction of a novel symbolic representation method: Trend Feature Symbolic Approximation (TFSA).
• Focus on retaining original time series trend features.
• Adaptation of the method for subsequent data mining tasks.

Main Results:

• TFSA effectively preserves key trend features of biomedical data series.
• The method is well-suited for data mining tasks like association rules mining.
• Experimental results demonstrate improved classification accuracy compared to existing methods.

Conclusions:

• TFSA offers a lower bounding guarantee for data representation.
• The proposed method enhances classification accuracy in biomedical data analysis.
• TFSA shows promise for advancing computer-assisted medical diagnosis.