Using a Decision Tree Algorithm Predictive Model for Sperm Count Assessment and Risk Factors in Health Screening...

Area of Science:

• Reproductive Biology
• Bioinformatics
• Public Health

Background:

• Infertility affects approximately 20% of couples globally.
• Sperm quality is a critical determinant of successful conception and artificial reproduction outcomes.
• Previous research utilized machine learning to identify ten risk factors for low sperm count in Taiwanese males.

Purpose of the Study:

• To develop a predictive model for healthy sperm counts using machine learning.
• To identify and analyze key risk factors influencing male fertility.
• To provide healthcare professionals with a tool for assessing male infertility.

Main Methods:

• Employed the Classification and Regression Trees (CART) algorithm to construct decision trees.
• Utilized ten identified risk factors from previous machine learning analysis.
• Evaluated decision tree performance using SMAPE, RAE, RRSE, and RMSE error metrics.

Main Results:

• The top-performing decision tree incorporated BMI, uric acid (UA), sleep duration (ST), total cholesterol/HDL-C ratio, and blood urea nitrogen (BUN).
• Confirmed the negative impact of metabolic syndrome indicators, such as high BMI, on sperm count.
• Highlighted the positive association between adequate sleep and male fertility, and suggested UA and T-Cho/HDL-C as novel factors.

Conclusions:

• A machine learning-based predictive model was established to assess low sperm counts.
• Identified risk factors provide targets for future research and potential interventions.
• Further model refinement with additional data is recommended for enhanced accuracy.