International Journal of Innovation Management and Organizational Behavior (IJIMOB)

Objective: The objective of this study was to develop and validate a machine learning–based predictive model for estimating innovation failure risk using organizational stress, workload distribution, and team conflict as primary predictors.

Methods and Materials: This quantitative cross-sectional study was conducted among 612 full-time employees from innovation-driven organizations in Malaysia. Data were collected using standardized survey instruments measuring organizational stress, workload distribution, team conflict, and perceived innovation failure risk. After psychometric validation, the dataset underwent preprocessing including normalization, outlier detection, and feature engineering. Innovation failure risk was converted into a binary classification outcome. Multiple machine learning classifiers were trained and compared, including logistic regression, support vector machines, random forest, gradient boosting, and extreme gradient boosting. Hyperparameter optimization and nested cross-validation were applied to ensure model stability and generalizability.

Findings: The XGBoost classifier achieved the highest predictive performance with an accuracy of 94%, precision of 93%, recall of 92%, F1-score of 92%, and AUC of 0.97, significantly outperforming all baseline models. Feature importance analysis revealed that emotional exhaustion and task overload were the strongest predictors of innovation failure risk, followed by relationship conflict and resource imbalance. The final model demonstrated high sensitivity for detecting high-risk innovation cases, confirming the robustness and reliability of the proposed predictive framework.

Conclusion: The findings demonstrate that innovation failure risk is strongly driven by human-centered organizational factors and can be accurately predicted using advanced machine learning models. The proposed framework provides organizations with a powerful early-warning system for preventing innovation breakdowns and strengthening innovation sustainability through proactive management of psychological and structural risk factors.

Objective: The objective of this study was to identify and quantify the relative importance of organizational citizenship behavior dimensions in predicting product innovation performance using an explainable machine learning framework.

Methods and Materials: This study employed a quantitative cross-sectional design with data collected from 412 employees working in medium and large organizations. Organizational citizenship behaviors were measured across five dimensions including conscientiousness, civic virtue, courtesy, altruism, and sportsmanship, while product innovation performance was assessed using multidimensional performance indicators. Data were analyzed using the XGBoost algorithm with hyperparameter optimization and five-fold cross-validation. Feature importance was examined using both gain-based metrics and SHAP (SHapley Additive exPlanations) values to ensure transparent and robust interpretation of predictor contributions. Model robustness was evaluated through comparison with Random Forest and Elastic Net models.

Findings: The XGBoost model demonstrated strong predictive performance, explaining 68.4% of the variance in product innovation performance. Feature importance analysis revealed that conscientiousness was the most influential predictor, followed by civic virtue, courtesy, altruism, and sportsmanship. Conscientiousness exhibited the highest SHAP contribution, indicating its dominant role in driving innovation outcomes. Comparative analysis showed that XGBoost significantly outperformed Random Forest and Elastic Net across all evaluation metrics, confirming its superior capacity to capture nonlinear behavioral effects on innovation performance.

Objective: The objective of this study was to develop and validate a graph neural network model for predicting open innovation success based on trust networks, communication density, and collaborative behaviors in Nigerian organizations.

Methods and Materials: This study employed a quantitative, cross-sectional research design involving 487 professionals from innovation-intensive organizations in Nigeria across the technology, manufacturing, finance, telecommunications, and energy sectors. Data were collected using validated instruments measuring inter-organizational trust, communication density, and collaborative behaviors, combined with objective network data extracted from organizational collaboration platforms. Multilayer networks were constructed in which nodes represented individuals and edges represented trust relations, communication ties, and collaborative interactions. A graph neural network architecture integrating graph convolutional and attention mechanisms was trained to predict open innovation success. Model performance was evaluated using root mean squared error, mean absolute error, and explained variance, and results were compared against baseline machine learning models.

Findings: The graph neural network demonstrated strong predictive performance (R² = 0.82, RMSE = 0.31, MAE = 0.24), significantly outperforming gradient boosting (R² = 0.64), random forest (R² = 0.61), and support vector regression (R² = 0.58). Explainability analysis revealed that trust networks exerted the strongest influence on innovation success (importance weight = 0.41), followed by collaborative behaviors (0.30) and communication density (0.29). Multilevel network analysis showed that team-level structures had the largest standardized effect on innovation success (β = 0.44, p < 0.001), followed by individual-level (β = 0.36, p < 0.001) and organizational-level networks (β = 0.28, p < 0.001).

Conclusion: The findings demonstrate that open innovation success is primarily driven by the structure and quality of relational networks, and that graph neural networks provide a powerful and superior framework for modeling innovation outcomes within complex organizational ecosystems.

Objective: The objective of this study was to model organizational ambidexterity by applying ensemble machine learning techniques to identify the behavioral and structural predictors of exploratory and exploitative innovation.

Methods and Materials: This explanatory study employed a cross-sectional design involving 487 middle- and senior-level managers from medium and large organizations across major industries in Chile. Data were collected using validated instruments measuring leadership cognitive flexibility, learning orientation, psychological safety, risk tolerance, cross-functional integration, decentralization, resource flexibility, knowledge-sharing systems, and dual innovation outcomes. The analytical framework integrated traditional statistical validation with an ensemble learning architecture composed of Random Forest, Gradient Boosting, XGBoost, and Support Vector Regression models. Model training applied stratified sampling, five-fold cross-validation, and hyperparameter optimization, while performance was evaluated using R², RMSE, MAE, and explained variance. Explainable AI techniques based on SHAP were employed to interpret nonlinear relationships and predictor contributions.

Findings: The ensemble model demonstrated superior predictive performance for both exploratory innovation (R² = 0.81, RMSE = 0.25) and exploitative innovation (R² = 0.84, RMSE = 0.22), significantly outperforming individual machine learning algorithms. Leadership cognitive flexibility and learning orientation emerged as the strongest predictors of exploratory innovation, whereas cross-functional integration and structural decentralization exerted the greatest influence on exploitative innovation. Psychological safety, risk tolerance, knowledge sharing, and resource flexibility contributed significantly to both innovation dimensions, with SHAP analysis revealing asymmetric and nonlinear interaction effects across predictors.

Conclusion: The results confirm that organizational ambidexterity is a systemic, nonlinear phenomenon driven by the dynamic interaction of behavioral and structural factors and that ensemble learning provides a powerful methodological approach for modeling this complexity, offering both theoretical advancement and practical guidance for innovation management.

Objective: The objective of this study was to develop and validate a machine learning–based behavioral analytics framework for identifying organizational innovation bottlenecks through the interaction of leadership, psychological, and behavioral factors.

Methods and Materials: This study employed a cross-sectional explanatory design involving 547 employees and middle-level managers from diverse organizations in Georgia. Data were collected using validated behavioral, psychological, and organizational measures capturing resistance to change, psychological safety, leadership support, communication friction, knowledge sharing, learning orientation, and innovation outcomes. Organizational performance indicators were integrated with survey data to enhance behavioral signal extraction. Gradient boosting algorithms (XGBoost, LightGBM, CatBoost) and an optimized ensemble model were implemented using five-fold cross-validation and Bayesian hyperparameter tuning. Feature engineering and explainable artificial intelligence techniques (SHAP values) were applied to uncover the relative importance and interaction effects of predictors.

Findings: The ensemble model demonstrated strong predictive performance (R² = 0.846, RMSE = 0.387), explaining nearly 85% of the variance in innovation bottleneck intensity. Resistance to change was the strongest positive predictor, while leadership support for innovation, psychological safety, knowledge sharing quality, and proactive behavior significantly reduced bottleneck severity. Communication friction and excessive process formalization amplified innovation constraints. Behavioral segmentation revealed four distinct innovation profiles, with “Resistant Traditionalists” exhibiting the highest bottleneck levels and “Adaptive Innovators” the lowest.

Conclusion: The findings confirm that innovation bottlenecks are systemic behavioral–organizational phenomena emerging from complex non-linear interactions among leadership dynamics, employee psychology, communication structures, and organizational culture.

Objective: The objective of this study was to develop and validate a deep learning model capable of explaining and predicting the combined effects of transformational leadership and organizational climate on continuous innovation capability within organizational settings.

Methods and Materials: The study employed a quantitative cross-sectional design using data collected from 684 full-time employees and managers working in medium-to-large organizations in Germany. Transformational leadership was measured using the Multifactor Leadership Questionnaire, organizational climate was assessed through a standardized organizational climate scale, and continuous innovation capability was captured using a multidimensional innovation capability instrument. After establishing construct validity and reliability through confirmatory factor analysis, the dataset was analyzed using a deep neural network implemented with TensorFlow and Keras. Data were split into training, validation, and test sets (70–15–15). Model optimization employed Bayesian hyperparameter tuning, and performance was evaluated using mean squared error, root mean squared error, coefficient of determination, and prediction accuracy. Model interpretability was enhanced using SHAP analysis and permutation feature importance techniques.

Findings: Inferential results demonstrated strong predictive performance of the deep learning model (R² = .89 training, .87 validation, .86 test; prediction accuracy = 90.4% on test set). Transformational leadership and organizational climate jointly explained a substantial proportion of variance in continuous innovation capability. Feature importance analysis revealed inspirational motivation and innovation-oriented climate as the most influential predictors, followed by intellectual stimulation and supportive climate. The deep neural network significantly outperformed random forest, gradient boosting, and multiple regression models.

Objective: The present study aimed to investigate the threshold effects of gross fixed capital formation on the relationship between healthcare infrastructure and the life expectancy index in Iran.

Methods and Materials: This applied and descriptive–analytical study employed an ex post facto design using annual time-series data for Iran covering the period 1986–2023. The empirical analysis was conducted using EViews software. To examine nonlinear dynamics and potential regime-switching behavior, a first-order Logistic Smooth Transition Regression (LSTR1) model was estimated, with gross fixed capital formation serving as the transition variable. Prior to model estimation, Phillips–Perron unit root tests were applied to assess stationarity, and the Johansen–Juselius cointegration approach was used to test for long-run equilibrium relationships among the variables. The model incorporated healthcare infrastructure indicators, including hospital beds, physicians, nurses and midwives, public and private health sector investment, labor force participation, population growth, and literacy rate. Diagnostic tests were performed to assess autocorrelation, heteroskedasticity, parameter stability, and remaining nonlinearity.

Findings: The results indicate that all variables are integrated of order one and cointegrated in the long run. The linearity test rejects the null hypothesis of linearity, confirming the appropriateness of the LSTR1 specification. Gross fixed capital formation exhibits a statistically significant threshold value of −0.211, dividing the model into two distinct regimes. In the nonlinear (high-capital) regime, the coefficients of gross fixed capital formation, hospital beds, physicians, nurses and midwives, and both public and private health sector investment are positive and statistically significant, with greater magnitudes compared to the low-capital regime. Diagnostic tests confirm the absence of autocorrelation and heteroskedasticity, and parameter constancy tests support regime-dependent coefficient variation. The adjusted R² of 0.88 indicates strong explanatory power.

Conclusion: The findings demonstrate that the effectiveness of healthcare infrastructure in improving life expectancy in Iran is conditional upon the level of gross fixed capital formation. When capital accumulation surpasses a critical threshold, the impact of healthcare infrastructure intensifies significantly, highlighting the importance of sustained macroeconomic investment in enhancing health outcomes.

Objective: The objective of this study was to predict consumers’ purchase intention based on salesperson behavior, economic utility, social utility, and psychological utility.

Methods and Materials: This applied study employed a descriptive–correlational design and was conducted using a cross-sectional survey method. The statistical population consisted of consumers who completed online questionnaires during May and June 2025. A total of 377 valid questionnaires were analyzed after excluding incomplete responses. Data collection instruments included researcher-developed questionnaires measuring purchase intention, salesperson behavior, economic utility, social utility, and psychological utility, all assessed on five-point Likert scales. Content validity was confirmed by subject-matter experts, and reliability was evaluated using Cronbach’s alpha. Data analysis was carried out using SPSS version 26, applying Pearson correlation analysis and stepwise multiple regression to examine the predictive relationships among variables.

Findings: Inferential analyses indicated that purchase intention was positively and significantly correlated with salesperson behavior, economic utility, social utility, and psychological utility (p ≤ .01). Stepwise multiple regression analysis showed that economic utility entered the regression model in the first step and explained 24% of the variance in purchase intention (p = .001). In the second step, social utility entered the model and, together with economic utility, explained 27% of the variance in purchase intention (p = .001). Psychological utility and salesperson behavior did not retain significant predictive power in the final regression model.

Conclusion: The findings demonstrate that consumers’ purchase intention is primarily driven by economic utility and, to a lesser extent, social utility, highlighting the dominant role of cost–benefit evaluations and social value considerations in purchase decision-making. These results suggest that while psychological benefits and salesperson behavior are important correlates, their influence on purchase intention may be indirect and mediated through perceived value dimensions.