@article{4790,
  author = {Nguyen Minh Tuan},
  title = {Exploratory Factor Analysis and Random Forest Modeling of ROI Determinants in Robotic Process Automation (RPA) Implementation},
  journal = {Journal of Intelligent Computing},
  year = {2026},
  volume = {17},
  number = {3},
  doi = {https://doi.org/10.6025/jic/2026/17/3/122-140},
  url = {https://www.dline.info/jic/fulltext/v17n3/jicv17n3_2.pdf},
  abstract = {This study investigates the determinants of Return on Investment (ROI) in enterprise Robotic Process
Automation (RPA) implementations through an integrated analytical framework combining Exploratory
Factor Analysis (EFA) and Random Forest regression modeling. Using project-level data comprising five key
performance indicators Robots Deployed, Budget (USD), Annual Savings (USD), ROI (%), and Employee
Hours Saved the research identifies underlying latent structures and evaluates the predictive power of
financial and operational variables on automation returns. EFA revealed a two-factor solution explaining
72.71% of total variance, interpreted as Financial Performance and Operational Efficiency. The Random
Forest model demonstrated excellent predictive accuracy (Mean R² = 0.962, MAE = 7.90) with stable crossvalidation
performance (SD = 0.0055). Feature importance analysis identified Annual Savings (35.91%),
Budget (32.13%), and Employee Hours Saved (27.29%) as the dominant ROI predictors, while Robots Deployed
contributed only 4.68%. SHAP-based explainability confirmed that realized economic benefits and workforce
productivity gains constitute the primary mechanisms driving ROI generation. Although the Kaiser-Meyer-
Olkin measure (0.412) indicated limited common variance, Bartlett's Test of Sphericity was significant (² =
117,907.48, p < .001), supporting factor extraction. The findings suggest that successful RPA implementations
should prioritize measurable financial outcomes and operational efficiency over automation scale expansion.
This research contributes empirical evidence on RPA value creation and provides actionable guidance for
organizations seeking to optimize automation investments. Conceptual evaluation of XGBoost indicated
only marginal expected improvements, confirming the robustness of the ensemble learning approach.},
}