Predicting Employee Innovative Work Behavior from Psychological Safety and Proactive Personality Using Machine Learning Models

Abstract

Objective: This study aims to predict employee innovative work behavior by evaluating the distinct and interactive contributions of psychological safety and proactive personality through the application of advanced machine learning algorithms.

Methods and Materials: A cross-sectional, quantitative design was utilized, drawing a stratified random sample of full-time employees from various dynamic corporate sectors across Indonesia. Data were collected via self-administered, translated questionnaires utilizing established Likert-scale instruments. To evaluate the predictive relationships, the dataset was partitioned into an training and testing split. Four machine learning models—Multiple Linear Regression, Support Vector Regression, Random Forest Regressor, and Gradient Boosting Regressor—were trained and evaluated using five-fold cross-validation, while SHapley Additive exPlanations (SHAP) values were computed to interpret feature importance and non-linear interactions.

Findings: The non-linear ensemble models outperformed traditional linear approaches, with the Gradient Boosting Regressor demonstrating the highest predictive accuracy on the testing set ( , , ). Feature importance analysis using SHAP values identified proactive personality as the dominant predictor of innovative work behavior (mean absolute SHAP value = ), closely followed by psychological safety (mean absolute SHAP value = ). Furthermore, the dependency plots revealed a critical non-linear interaction, indicating that high levels of proactive personality only consistently translate into innovative behavior when a baseline threshold of psychological safety is present in the work environment.

Conclusion: Organizations must concurrently hire for proactive traits and cultivate psychologically safe climates, as both are fundamentally intertwined in actualizing and predicting employee innovation.