Cost-Sensitive KNN Algorithm for Cancer Prediction Based on Entropy Analysis.

Area of Science:

• Biochemistry
• Medical Diagnostics
• Bioinformatics

Background:

• Early cancer diagnosis is crucial for effective treatment planning, improving survival rates and patient quality of life.
• Current methods like imaging and needle biopsy have limitations in early detection and can be invasive.
• Analyzing blood protein indexes offers a less invasive approach to monitor health status and detect cancer early.

Purpose of the Study:

• To investigate the potential of serum protein marker entropy analysis for early cancer detection.
• To develop and validate a cost-sensitive analysis model for improved cancer recognition accuracy.
• To explore the differences in serum protein marker complexity between normal and cancer patients.

Main Methods:

• Analysis of entropy differences in 39 serum protein marker sequences across different patient groups.
• Development of a cost-sensitive analysis model to enhance cancer recognition.
• Validation of the model on an imbalanced dataset (897 instances: 799 normal, 44 liver cancer, 54 ovarian cancer).

Main Results:

• Significant differences in serum protein marker entropy were observed between different cancer types.
• Serum protein marker complexity was found to be higher in normal individuals compared to cancer patients.
• The developed model achieved a high accuracy of 95.21% despite dataset imbalance, with strong precision (0.807), recall (0.833), F1 (0.819), and AUC (0.92).

Conclusions:

• Serum protein marker entropy analysis shows promise for early cancer prediction.
• The cost-sensitive model effectively improves cancer recognition accuracy, offering a valuable tool for clinical application.
• This research provides a foundation for intelligent medical treatment and non-invasive cancer diagnostics.