Using Machine Learning to Identify the Most Powerful Self-Reported Predictors of Sexual Satisfaction

https://doi.org/10.1177/02654075211047004

Explanation

Sexual satisfaction is strongly linked to relationships and personal well-being, and previous research has found several individual (e.g., gender, self-esteem, attachment) and relationship (e.g., relationship satisfaction, relationship length, sexual desire) factors that predict sexual satisfaction. The goal of this study was to use modern machine learning to identify which variables are the strongest and weakest predictors of sexual satisfaction.

Previous research has relied on traditional statistical models, which are limited in their ability to estimate large numbers of predictors, nonlinear associations, and complex interactions, to predict sexual satisfaction across two samples through a machine learning algorithm, Random Forest (a potentially more flexible extension of the decision tree). The results were as follows (Total N = 1846; 754 individuals forming 377 couples were included in the study.)

We also used Sharpe Ray values, a game theory interpretation technique. This allowed us to estimate the size and direction of the impact of each predictor variable on the outcome of the model. The findings showed that sexual satisfaction is highly predictable with the relational variables (relationship satisfaction, importance of sex in the relationship, romantic love, and two-way desire) explaining the greatest variation in sexual satisfaction.