Predicting Parenting Inconsistency Using Machine Learning: Executive Dysfunction, Stress Reactivity, and Role Overload

Abstract

Objective: The present study aimed to predict parenting inconsistency using machine learning models based on executive dysfunction, stress reactivity, and role overload among parents.

Methods and Materials: This study employed a descriptive–correlational design with a machine learning predictive approach. The statistical population consisted of parents residing in Canada with at least one child aged 6 to 16 years, from which 412 participants were selected using stratified random sampling. Data were collected through standardized self-report instruments, including the Parenting Consistency Scale, the Behavior Rating Inventory of Executive Function–Adult Version (BRIEF-A), the Perceived Stress Reactivity Scale (PSRS), and the Role Overload Scale. Data preprocessing involved normalization, outlier detection, and handling missing values using multiple imputation. Statistical analyses were initially conducted using SPSS-27 to compute descriptive statistics and correlations. Subsequently, machine learning models including Random Forest, Support Vector Machine, Gradient Boosting, and Artificial Neural Networks were implemented using Python-based libraries. Model performance was evaluated using accuracy, mean squared error (MSE), coefficient of determination (R²), and area under the receiver operating characteristic curve (AUC-ROC), with a 70/30 train–test split and cross-validation procedures applied.

Findings: The results indicated that executive dysfunction, stress reactivity, and role overload were all significant positive predictors of parenting inconsistency. Among these variables, executive dysfunction demonstrated the strongest predictive contribution across all models. Machine learning analyses revealed that the Gradient Boosting model achieved the highest performance (Accuracy = 0.86, R² = 0.66, AUC = 0.90), followed by Random Forest (Accuracy = 0.84, R² = 0.62). Artificial Neural Networks showed moderate predictive performance, while Support Vector Machine exhibited comparatively lower explanatory power. Feature importance analyses confirmed that executive dysfunction was the most influential predictor, followed by role overload and stress reactivity.

Conclusion: Integrating executive functioning, stress regulation, and environmental demands within predictive frameworks offers a comprehensive understanding of parenting variability and provides a foundation for developing targeted interventions to promote consistent parenting practices.