Utilizing Ensemble Meta-Learning Based Stacking Deep Learning for Predicting Digital Transformation Success by Leveraging Organizational Agility

Abstract

The aim of this study was to develop a machine learning-based framework for predicting the success of digital transformation in organizations by emphasizing the role of organizational agility and introducing a quantitative index for measuring digital readiness. This study was conducted using a standard digital transformation dataset containing organizational and technological variables such as company size, adopted digital tools, digital skills training, customer engagement changes, and organizational performance indicators. First, an Organizational Agility Index (OAI) was developed through a linear combination of several key organizational variables to quantitatively measure organizational readiness and flexibility toward digital transformation. Then, the K-Means clustering algorithm was employed to classify organizations into three levels of digital transformation success—low, medium, and high—based on the impact of digital transformation on revenue and efficiency. During preprocessing, categorical variables were encoded using One-Hot Encoding, while numerical features were normalized through standardization techniques. Subsequently, a stacking deep learning architecture consisting of MLP, RNN, LSTM, CNN, and ResMLP models was implemented as base learners, and their outputs were integrated using a Random Forest meta-learner. Model performance was evaluated using accuracy, sensitivity, precision, F-score, MAE, and RMSE metrics, and the training-testing process was repeated across 30 independent runs. The findings demonstrated that the proposed stacking deep learning framework significantly outperformed individual models. The Random Forest meta-learner achieved an accuracy of 96.86%, representing the highest predictive performance among all compared approaches. Sensitivity, precision, and F-score values were also considerably high, while the variance across repeated executions remained minimal, indicating strong stability and generalizability of the proposed framework. The results further revealed that integrating the Organizational Agility Index with advanced deep learning and ensemble meta-learning architectures substantially enhanced the prediction capability of digital transformation success. Moreover, combining multiple neural architectures enabled the extraction of more complex hidden patterns from organizational data and reduced prediction errors compared with single-model approaches. The results of this study indicate that stacking deep learning frameworks integrated with organizational indicators can provide an effective approach for predicting digital transformation success. The proposed Organizational Agility Index enabled quantitative assessment of digital readiness, while its integration with deep learning models improved prediction accuracy and stability. The proposed architecture demonstrated a strong capability to identify complex patterns within organizational data and can support managers and policymakers in strategic decision-making, resource allocation, and evaluating organizational preparedness for digital transformation. Furthermore, the framework has the potential to be implemented in real-world organizational environments as an intelligent predictive system for assessing digital transformation outcomes.