Integrating Gradient Boosting and Artificial Intelligence Agent Based Workflow Automation for Employee Attrition Prediction and Human Resource Meeting Scheduling

The main limitations are as follows. The study relies on one public, fictional dataset whose age, provenance and representativeness are limited, so generalisation to real organisations is limited. Class imbalance is handled only through a stratified split, without resampling or class weights, and 21 of the 47 test leavers are missed at the 0.50 threshold. No performance is reported at the 0.80 scheduling cutoff. There is no formal explainability analysis such as SHAP [8], and the method that produces the top reasons in the payload has not been evaluated. Evaluation relies on one split with 47 positive test records. The test bench consists of 20 runs on synthetic data, with no load testing and no failure injection for retries. The invitation description carries risk drivers. There is no external validation. Employee-level scoring carries privacy and ethical risks [9]. The system depends on external APIs and on workflow reliability. Finally, there is no evidence that scheduling meetings reduces attrition. Future work includes validation on appropriately authorised organisational data; comparison with other classifiers under the same protocol, using cross- validation with confidence intervals; threshold tuning and class-imbalance treatment inside training folds, with precision and recall reported at each risk-tier cutoff; calibration and explainability analysis; fairness evaluation across groups; minimisation of the payload and the invitation description; load testing and failure injection for the workflow and its retries; and a controlled study of whether the workflow helps HR staff. The items listed here are not part of the implemented system.

[1]               A. Raza, K. Munir, M. Almutairi, F. Younas, and M. M.S. Fareed, “Predicting employee attrition using machine learning approaches,” Appl. Sci., vol. 12, no. 13, Art. no. 6424, 2022, doi: 10.3390/app12136424. [2]               W. Li, “A transformer-based deep learning framework to predict employee attrition,” PeerJ Comput. Sci., 2023, doi: 10.7717/peerj-cs.1570. [3]               M. Nandal, V. Grover, D. Sahu, and M. Dogra, “Employee attrition: Analysis of data-driven models,” EAI Endorsed Trans. Internet Things, vol. 10, no. 1, 2024, doi: 10.4108/eetiot.4762. [4]               “IBM HR analytics employee attrition & performance,” Kaggle dataset (contributor: pavansubhasht). [Online]. Available: https://www.kaggle.com/datasets/pavansubhasht/ibm-hr-analytics-attrition-dataset (accessed Oct. 11, 2026). [5]               J. H. Friedman, “Greedy function approximation: A gradient boosting machine,” Ann. Statist., vol. 29, no. 5,pp. 1189–1232, 2001, doi: 10.1214/aos/1013203451. [6]               F. Pedregosa et al., “Scikit-learn: Machine learning in Python,” J. Mach. Learn. Res., vol. 12, pp. 2825–2830, 2011. [7]               T. Saito and M. Rehmsmeier, “The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets,” PLoS ONE, vol. 10, no. 3, Art. no. e0118432, 2015, doi: 10.1371/journal.pone.0118432. [8]               S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems, vol. 30, 2017, pp. 4765–4774. [9]               N. Mehrabi, F. Morstatter, N. Saxena, K. Lerman, andA. Galstyan, “A survey on bias and fairness in machine learning,” ACM Comput. Surv., vol. 54, no. 6, pp. 1–35, 2021, doi: 10.1145/3457607. [10]             n8n, “Google Calendar node,” n8n Documentation. [Online]. Available: https://docs.n8n.io/integrations/builtin/app-nodes/n8n-nodes-base.googlecalendar/ (accessed Oct. 11, 2026).