Enhancing the efficacy of depression detection system using optimal feature selection from EHR.

Area of Science:

• Computational psychiatry
• Machine learning in healthcare

Background:

• Early depression diagnosis is critical but relies heavily on clinician expertise.
• Developing automated tools can support and enhance diagnostic accuracy.

Purpose of the Study:

• To create an automated system for assisting depression diagnosis using machine learning.
• To improve the performance and efficiency of depression detection models.

Main Methods:

• Utilized a dataset of 4184 individuals, including biometric and demographic data.
• Employed machine learning classifiers: Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost).
• Integrated feature selection algorithms (RFE, MI, PSO, GA, FA) with majority voting to reduce feature set size by 45.90% (from 61 to 33).

Main Results:

• The enhanced models achieved classification accuracies ranging from 84.18% to 88.46%.
• This represents a significant performance improvement compared to pre-existing models (83.76-85.89%).
• The feature selection methods enhanced both the performance and computational efficiency of the diagnostic process.

Conclusions:

• The proposed machine learning models with feature selection significantly improve depression diagnosis accuracy and efficiency.
• Automated tools can effectively assist clinicians in the early and accurate diagnosis of depression.
• This approach offers a promising direction for developing advanced computational diagnostic aids in mental health.