Seminal quality prediction using data mining methods.

Area of Science:

• Reproductive Health
• Data Mining Applications
• Artificial Intelligence in Medicine

Background:

• Lifestyle diseases are increasing due to altered habits like alcohol consumption and smoking.
• Male fertility rates, specifically sperm quantity, have declined significantly over the past two decades.
• Both lifestyle and environmental factors are implicated in reduced semen quality.

Purpose of the Study:

• To identify lifestyle and environmental factors influencing male seminal quality and fertility.
• To apply data mining methods for predicting fertility status based on these factors.

Main Methods:

• Five artificial intelligence techniques were employed: Multilayer Perceptron (MLP), Decision Tree (DT), Navie Bayes (Kernel), Support Vector Machine (SVM), and SVM+PSO.
• Eight feature selection techniques were utilized to identify the most relevant predictors of seminal quality.
• These methods were applied to a fertility dataset comprising 100 instances and nine attributes.

Main Results:

• SVM+PSO achieved the highest accuracy (94%) and Area Under Curve (AUC) (0.932) in predicting seminal quality.
• Feature selection identified age, season, surgical intervention, alcohol consumption, smoking, hours spent sitting, and accidents as significant factors.
• Feature selection methods demonstrably improved the predictive accuracy of all applied AI techniques.

Conclusions:

• Data mining techniques offer a viable alternative to traditional medical tests for predicting fertility status.
• SVM+PSO demonstrated superior performance in predicting male fertility compared to other tested methods.
• The study highlights the importance of lifestyle and environmental factors in male reproductive health.