Support Vector Machine Classification of Employees with High Innovative Work Behavior Using Psychological and Organizational Indicators

Abstract

Objective: This study aimed to classify Tunisian employees with high innovative work behavior using a Support Vector Machine model based on psychological and organizational indicators and to determine the relative contribution of these predictors to classification performance.

Methods and Materials: This quantitative cross-sectional predictive study was conducted among 418 employees working in public and private organizations in Tunis, Sfax, Sousse, and Monastir, Tunisia. Participants completed standardized measures of innovative work behavior, psychological capital, work engagement, intrinsic work motivation, creative self-efficacy, perceived organizational support, perceived supervisor support, organizational climate for innovation, job autonomy, and knowledge sharing. Employees scoring at or above the 75th percentile of innovative work behavior were classified as having high innovative work behavior. The dataset was divided into stratified training and independent test sets using an 80/20 ratio. A radial basis function Support Vector Machine was optimized through stratified five-fold cross-validation and grid search for C and gamma parameters. Classification performance was evaluated using accuracy, sensitivity, specificity, precision, F1 score, balanced accuracy, and ROC-AUC. Permutation-based feature importance was used to estimate the relative predictive contribution of each indicator.

Findings: The optimized SVM achieved cross-validated accuracy of .871, balanced accuracy of .858, and ROC-AUC of .918. In the independent test set, the model achieved accuracy of .857, sensitivity of .810, specificity of .873, precision of .680, F1 score of .739, balanced accuracy of .841, and ROC-AUC of .910. Creative self-efficacy showed the greatest permutation importance (.082), followed by organizational climate for innovation (.069), psychological capital (.058), work engagement (.046), knowledge sharing (.040), job autonomy (.034), intrinsic work motivation (.029), perceived organizational support (.024), and perceived supervisor support (.017).

Conclusion: Psychological and organizational indicators can classify employees with high innovative work behavior with substantial out-of-sample accuracy, with creative self-efficacy and organizational climate for innovation emerging as the strongest discriminative predictors.