Predictive Models of Life Satisfaction in Older People: A Machine Learning Approach.

Area of Science:

• Gerontology
• Psychology
• Data Science

Background:

• Traditional studies on older adult life satisfaction often rely on simplistic linear models.
• Real-world relationships influencing well-being are frequently complex and non-linear.

Purpose of the Study:

• To identify key predictors of life satisfaction in adults aged 50 and above.
• To compare the predictive power of various machine learning models against traditional statistical methods.

Main Methods:

• Utilized a large dataset (n=34,630) from the RAND Health and Retirement Study.
• Employed machine learning techniques including Support Vector Regression (SVR), Multiple Linear Regression (MLR), Ridge Regression (RR), LASSO, K Nearest Neighbors (KNN), and Decision Tree Regression (DT).

Main Results:

• Subjective social status, positive emotions, and negative emotions emerged as the most significant predictors.
• The Support Vector Regression (SVR) model demonstrated the highest accuracy in predicting life satisfaction.
• While KNN and DT showed good model fitting, SVR excelled in validation and generalization.

Conclusions:

• Machine learning approaches offer superior predictive capabilities for understanding life satisfaction in older adults compared to simple correlation.
• Identifying key psychosocial factors is crucial for enhancing well-being in the aging population.