Data Intelligence for Workforce Management: Machine Learning Models for Payroll and Resource Optimization

Authors

• Raghavendra Sunku Data engineer

Abstract

This study provides a comprehensive analysis of team pay analytics and resource optimization platforms designed for large-scale operational organizations in the aviation, utilities, and field services industries. The research examines the implementation of data-driven workforce management systems that integrate scheduling algorithms, compensation modeling, and performance analytics to optimize labor allocation while ensuring fair compensation practices. Using machine learning approaches, specifically gradient boosting regression and XGBoost regression, the study analyzes employment data from 22 employees across multiple dimensions, including hours worked, skill levels, shift types, overtime hours, and weekly compensation.The method uses team learning techniques to predict weekly pay based on employee characteristics, with both algorithms demonstrating exceptional training performance, achieving R² values of 0.9997. However, the pilot phase results revealed significant performance degradation, with R² values dropping to approximately 0.90, indicating overfitting challenges common in team learning applications. XGBoost showed slightly superior generalization capabilities with better error control and built-in regularization algorithms compared to standard gradient boosting.The platform addresses important challenges in workforce management, including complex scheduling requirements, regulatory compliance, variable pay structures, and team-based productivity optimization. Key findings reveal strong positive correlations (0.91-0.98) between working hours, skill levels, overtime hours, and weekly pay, suggesting merit-based pay systems. While both machine learning models show promise for salary prediction with 90% accuracy on unobserved data, the research demonstrates that proper validation and regularization strategies are essential for practical use in human resource analytics and compensation planning systems.

Key Words: Team Payroll Analysis, Resource Optimization, Team Management, Machine Learning Regression, Compensation Modeling, XGBoost Algorithm, Team Learning, Predictive Analytics.

INTRODUCTION

In large-scale operational organizations such as aviation, public utilities, or field service teams, a team payroll analytics and resource optimization platform that integrates workforce planning, compensation planning, and performance has become a critical tool. The system uses data-driven insights to strategically allocate workers, control labor costs, and increase productivity by ensuring that each task is matched to the most appropriate team member at the right time, with fair and transparent compensation. At the heart of this system is a combination of real-time analytics, intelligent scheduling algorithms, and payroll modeling tools. Successful implementation of such a platform requires careful balancing of multiple interdependent factors, including labor laws, employee preferences, skill distribution, fatigue mitigation, and budget constraints, while maintaining service quality and safety.[1]One of the most demanding aspects of team management is the complexity of the scheduling process. In industries such as aviation, energy distribution, and shipping, creating effective schedules involves navigating complex regulatory requirements, union contracts, mandatory rest periods, and unpredictable operational variables. For example, in airline operations, crew members must be assigned to work sequences or “pairings” that follow strict guidelines regarding work hours, rest periods, and certifications. [2] To handle this, planners increasingly rely on sophisticated optimization techniques such as linear and integer programming.

These methods allow them to evaluate a wide range of combinations of crew tasks to identify cost-effective and regulatory-compliant schedules. A well-integrated crew pay platform takes this a step further by linking each schedule pairing to its associated compensation, providing visibility into the financial implications and aiding better decision-making.[3]The financial aspects of team management are equally important. Derived from the workforce economy, compensation systems must not only motivate workers but also ensure internal fairness and alignment with market standards. Traditional fixed pay models are gradually being replaced by more nuanced structures that include performance-based bonuses, skill-level differentials, and overtime incentives.[4] This shift reflects a broader effort to align employee rewards with organizational goals. The platform must be able to model and manage variable pay structures across diverse teams whose skill levels, training, and work output vary significantly. This ensures that highly skilled or talented workers are appropriately compensated while avoiding inefficiency and dissatisfaction among peers.[5]Beyond individual compensation, modern workforce strategies emphasize team-based productivity. Studies of team incentives and workforce diversity show that thoughtfully structured teams—comprised of individuals with varying levels of expertise—can achieve better results, especially when structured in a way that fosters collaboration and mutual learning. In team-based environments, such as maintenance units or flight crews, it is essential to design both the team structure and the reward mechanism accurately. This platform can measure team output through metrics such as completion times, service quality, and incident reduction, and then distribute team-based bonuses accordingly.

This requires a tightly integrated system that combines attendance records, work progress, and human resource data into a single analytical interface.[6]Another key strength of such a platform lies in its ability to reduce excessive overtime and inefficiencies in scheduling. As seen in the case of application team management, reactive scheduling practices can lead to disproportionate overtime costs. Implementing an optimization tool such as the Resource Allocation and Scheduling Tool (RAPT) can address this issue by predicting fluctuations in workload and integrating emergency situations into the planning phase. [7] A team pay analytics system should have similar capabilities: dynamically recalculating team tasks and updating pay in response to changes such as weather delays, equipment outages, or staff absences. Simulating alternative scenarios helps ensure that schedules are efficient and cost-effective, even under uncertainty.[8]Transparency and fairness are equally important, both of which are key to building trust in compensation practices. A high-performance platform gives employees real-time access to see their hours worked, overtime accruals, and bonus eligibility. It also enables feedback mechanisms that can quickly resolve errors or disputes. [9] For supervisors, the system provides performance dashboards to help them track trends, identify outliers, and maintain compliance with both company policy and legal regulations.

This is especially important in highly regulated industries like healthcare and aviation, where missteps can have serious operational and legal consequences.[10]In addition to real-time scheduling and payroll management, the platform also serves as a powerful predictive tool for workforce planning. By analyzing historical data on absenteeism, performance fluctuations, or high-risk shifts (such as overnight work), the system can provide actionable insights to reduce error rates, fatigue, and turnover. Machine learning models can detect hidden trends in payroll complaints, productivity drops, or scheduling conflicts that might otherwise go unnoticed by human planners.[11] One of the most valuable functions of such platforms is their ability to support cost efficiency without compromising quality. Case studies from industries such as offshore shipping operations illustrate how comprehensive systems that combine enterprise resource planning (ERP), team tracking, and financial analytics can provide real-time insights into labor-related costs. [12] Centralizing data on team availability, work scheduling, purchasing, and payroll within a unified analytics environment reduces duplication and streamlines strategic decisions. For maximum efficiency, the Crew Pay Analytics platform should support integration with HR management systems, ERP software, and scheduling tools via APIs to ensure fluid data exchange and reduce administrative workload.[13]In the long term, the strategic advantage of a crew pay analytics and resource optimization platform lies in its ability to combine operational efficiency with employee morale and organizational resilience. In times of increased demand or budget constraints, organizations with such tools can dynamically adapt their workforce models while preserving equity, satisfaction, and productivity among their workforce. [14] As digital transformation continues to reshape the modern workplace, such platforms are no longer optional—they are essential. They bring structure, accountability, and agility to a space once governed by fragmented processes and manual oversight. [15]

Method

Gradient Boosting Regression and XGBoost Regression represent two state-of-the-art ensemble learning approaches that have demonstrated exceptional performance in predicting weekly pay based on employee characteristics. Both algorithms belong to the boosting family of machine learning techniques, which sequentially build multiple weak learners to build a robust predictive model. However, their implementation strategies, optimization techniques, and performance characteristics reveal important differences that affect their practical use in compensation analysis. Gradient Boosting Regression works through an iterative process, where each new model attempts to correct the errors of its predecessors.

The algorithm starts with a simple predictor, typically the mean of the target variable, and then adds a series of models that focus on minimizing the residual errors. Each iteration computes the slopes of the loss function and fits a new weak learner to these slopes, gradually improving the overall prediction accuracy. This continuous learning approach allows the model to capture complex nonlinear relationships between predictor variables such as working hours, skill levels, overtime hours, and shift types. The strength of the algorithm lies in its ability to handle mixed data types and automatically detect feature correlations without extensive preprocessing. XGBoost (Extreme Gradient Boosting) refers to an optimized and scalable implementation of gradient boosting that includes several advanced features to improve performance and prevent overfitting. The algorithm introduces regularization terms in its objective function, including both L1 and L2 penalties that control the model complexity. XGBoost uses a highly sophisticated tree construction process using second-order derivatives, which enables more accurate optimization steps. In addition, it implements advanced techniques such as column sub sampling, row sub sampling, and early stopping algorithms that improve generalization capabilities. The algorithm's parallel processing capabilities and memory optimization make it computationally efficient for large datasets. The training performance analysis reveals that both models achieved nearly identical exceptional results on the weekly wage prediction task. Both algorithms achieved R² values of 0.9997, indicating that they explained all the variation in the training data. The RMSE and MAE values were remarkably similar, with both models achieving prediction errors of approximately $2.19 and $1.15, respectively. This nearly perfect training performance demonstrates the ability of both algorithms to learn the complex relationships between employee characteristics and compensation structures. However, the gradient boosting model scored well on several metrics, including maximum error and mean absolute error, suggesting possible overfitting to training-based patterns.

The test phase performance revealed important differences in the generalization abilities between the two approaches. Both models experienced significant performance degradation from training to testing, with R² values decreasing to approximately 0.90. However, subtle differences emerged in their error patterns. XGBoost maintained slightly better control over prediction errors, with a maximum error of $80.00 compared to Gradient Boosting’s $93.44. The average absolute error favored XGBoost at $45.00, versus $33.19 for Gradient Boosting, although this comparison is complicated by the latter model’s correct training performance. The overfitting patterns observed in both models highlight a common challenge in ensemble learning applications. The dramatic performance gap between the training and testing phases indicates that both algorithms learned training-specific noise rather than common patterns. This is particularly evident in the correct training metrics achieved by Gradient Boosting, which indicates that the model memorized the training data rather than extracting underlying relationships. XGBoost’s built-in regularization mechanisms appear to provide some protection against overfitting, although not enough to completely eliminate the problem. For practical use in salary prediction systems, both models show promise, but their limitations should be carefully considered. R² values of 90% on test data indicate strong predictive ability, explaining most of the variation in employee compensation.

However, a significant increase in prediction errors on new data indicates the need for additional regularization techniques, cross-validation procedures, or hyper parameter optimization. XGBoost’s small advantage in generalization, combined with its computational efficiency and built-in regularization features, may make it the preferred choice for production environments. While both slope boosting and XGBoost excel at learning complex patterns in compensation data, the analysis reveals that their success depends heavily on proper model validation and regularization strategies. Future implementations should focus on techniques such as early stopping, feature selection, cross-validation, and ensemble methods to improve generalization while maintaining the strong predictive capabilities demonstrated in the training phases. Despite overfitting concerns, the models' ability to achieve 90% accuracy on unobserved data positions them as valuable tools for HR analysis and compensation planning when properly validated and used.

Materials

This statistical summary reveals key insights into an employee dataset of 22 employees, across multiple dimensions of their work patterns and pay. The data includes work hours, skill levels, shift types, overtime hours, and weekly pay, providing a comprehensive view of employment characteristics within this organization. The distribution of work hours shows that employees work an average of 42.3 hours per week, with relatively modest variation (standard deviation of 4.73 hours). This range spans from 35 to 50 hours per week, representing a mix of part-time and full-time positions. The 42-hour median closely matches the median, which represents a more normal distribution around stable full-time employment. The mid-range (38.25 to 45.75 hours) captures the majority of workers within the typical full-time parameters. The skill levels appear to follow a structured hierarchy, with minimum and maximum values given, likely rated on a scale of 1-5. The mean skill level of 3.05 indicates that most employees fall into the intermediate categories, with the mean exactly at 3.0, reinforcing this central tendency. The standard deviation of 1.25 indicates a fair skill diversity across the workforce, while the quartile distribution shows a fairly even spread from entry-level to advanced levels. The shift type data represents a binary classification system, with values of 1 and 2 representing different shift types (perhaps day/night or regular/alternating shifts). The mean of 1.45 and the median of 1.0 indicate that most employees work the primary shift type, with approximately 45% using the alternating shift system. Overtime hours present interesting patterns, with an average of 4.09 hours per week with considerable variation (standard deviation of 2.64). A range of 0 to 10 overtime hours indicates that some employees work strictly regular hours, while others consistently exceed standard schedules. A median of 4 hours indicates that overtime is fairly common across the workforce. Weekly wages show considerable wage variation, with an average of $862.73. The standard deviation of $177 indicates a significant wage spread, likely reflecting differences in skill levels and varying work hours. The median wage of $827.50 falls below the mean, indicating a slight rightward skew as a few higher-paid employees pull the mean upward. The median range ($711.25 to $987.50) covers most workers’ compensation, reflecting the company’s wage structure across a variety of skill levels and work patterns.

Analysis a nd Dissection

Table 1.Descriptive Statistics

This descriptive statistics table analyzes employment data for 22 workers, revealing important patterns in five key workplace variables. The dataset provides insights into work schedules, employee qualifications, shift arrangements, overtime practices, and compensation structures within this organization. Working hours show a typical full-time employment pattern, with employees working an average of 42.3 hours per week with a standard deviation of 4.73 hours. The distribution ranges from 35 to 50 hours, with the median (42 hours) closely matching the mean, indicating balanced work schedules. Most employees work between 38.25 and 45.75 hours based on the median range, suggesting stable full-time employment practices with some variation for operational flexibility. Skill levels operate on a scale of 1-5, with a mean of 3.05, with relatively low variance (standard deviation 1.25). The average skill level of 3.0 indicates that most workers have intermediate qualifications, while the quartile distribution shows a wide range of employees from entry-level (25th percentile at level 2) to advanced positions (75th percentile at level 4). This indicates a well-balanced group with a variety of expertise levels. Shift type appears to be binary, with values of 1 and 2 indicating different scheduling arrangements. The mean of 1.45 and the median of 1.0 indicate that approximately 55% of employees are in the primary shift type, while 45% work alternate shifts, indicating operational coverage over different periods. Overtime hours average 4.09 per week, with considerable variation (standard deviation 2.64). The range from 0 to 10 hours shows that some employees work strictly regular schedules, while others consistently exceed standard hours. The average overtime of 4 hours represents a common practice, indicating operational demands or employee willingness to work additional hours. Weekly wages show considerable variation, with an average of $625 to $1,200 with a mean of $862.73. The standard deviation of $177 reflects meaningful compensation differences, which may be related to skill levels and hours worked. The drop in average pay ($827.50) below the mean suggests that some of the highest-paid employees are creating a slight upward slope, indicating a compensation structure that rewards experience and expertise while maintaining competitive base wages at all skill levels.

FIGURE 1. Effect of Process Parameters

The pair wise distribution matrix reveals important relationships between workplace variables through both histograms and scatter plots. The diagonal histograms show that hours worked follow an approximate normal distribution centered around 40-42 hours, while skill levels show a more uniform distribution on a scale of 1-5. Shift type shows a clear binary pattern, with more employees in shift type 1. Overtime hours show a right-skewed distribution, with most employees working 2-6 overtime hours per week. Weekly wages show a relatively normal distribution with a slight right-skewed distribution, indicating that most employees earn between $700-1000 per week. The scatter plots reveal strong positive relationships between most variables, especially between hours worked, skill level, overtime hours, and weekly wages. More skilled employees tend to work longer hours and earn higher wages, suggesting a merit-based compensation system. The relationship between overtime hours and total compensation is particularly pronounced, indicating that overtime significantly affects earnings. Shift type shows weak correlations with other variables, suggesting that shift allocation may be more random or based on operational needs than employee characteristics.

FIGURE 2. Correlation heatmap

The correlation matrix measures the relationships found in the scatterplots, with correlation coefficients ranging from moderate to very strong. Hours worked, skill level, overtime hours worked, and weekly pay show exceptionally strong positive correlations (0.91-0.98), indicating that these variables are highly interdependent. This suggests that the company operates a performance-based system in which skilled employees work longer hours, accumulate more overtime, and receive proportionally higher compensation. Shift type shows weak correlations with the other variables (0.34-0.43), confirming that shift allocation operates relatively independently of employee skill levels or compensation structures. The strong correlation between skill level and weekly pay (0.98) indicates a well-structured pay scale that closely aligns compensation with employee qualifications and performance levels.

Table 2. Gradient Boosting Regression Weekly paytrain And Test Performance Metrics

The gradient boosting model exhibits exceptional training performance with an R² value of 0.9997, indicating a nearly perfect explanation of weekly wage variation. The training metrics demonstrate remarkable accuracy with $2.19 RMSE and $1.15 MAE, indicating minimal prediction errors. In particular, the model achieves perfect performance on the specified metrics, with the maximum error, MSLE, and mean absolute error all reaching 0.0000. This indicates that the algorithm has essentially memorized the training data and achieves flawless predictions on the salary dataset used for model development. The testing phase exhibits significant performance degradation, revealing clear overfitting issues. While still maintaining strong predictive power by explaining approximately 90% of the salary variation in the new data, the R² drops dramatically to 0.9006. The RMSE increases significantly to $49.26, while the MAE increases to $41.70, indicating significantly larger prediction errors in the missing data. The maximum error increases to $93.44, indicating a poor prediction deviation of the model, which is likely to occur at salary extremes where training examples were few. The explained variance score shows minimal degradation from 0.9997 to 0.9039, indicating that the model retains a reasonable ability to capture overall variation patterns. However, the mean absolute error increases from true zero to $33.19, meaning that half of all test predictions deviate by at least this amount, which represents an error of approximately 4% for typical salaries. The sharp difference between the correct training metrics and the degraded test performance clearly indicates overfitting, where the model has learned more training-specific noise than common salary prediction methods. Although the 90% R² on the test data shows that the model captures the underlying relationships between employee characteristics and compensation, the performance gap indicates that regularization techniques, feature selection, or ensemble methods are needed to improve generalization. The model shows promise, but refinement is needed before practical deployment in salary prediction systems.

FIGURE 3.Gradient Boosting Regression Weekly Pay Training

The training data results for the gradient boosting model demonstrate superior predictive performance with data points tightly clustered in the correct prediction order. The model successfully captures the relationship between input variables and weekly pay over the entire salary range from approximately $650 to $1120. Close alignment with the diagonal reference line indicates minimal prediction errors during training. The model appears to handle both low and high salary predictions accurately, indicating that it has effectively learned the underlying patterns in the compensation structure. The tight clustering around the reference line demonstrates the algorithm’s ability to capture complex nonlinear relationships between hours worked, skill levels, overtime, and resulting pay levels.

FIGURE 4.Gradient Boosting Regression Weekly Pay Testing

The test data evaluation reveals some relevant patterns in the generalization ability of the gradient boosting model. While the predictions for the middle salaries ($800-1000) are reasonably accurate, there are significant deviations at the extremes. The model shows less accuracy for lower-paid employees (around $700) and seems to have some difficulty with higher salary predictions. The scatter of points away from the reference line indicates a potential overfitting to the training data, indicating that the model may have learned training-specific patterns that do not generalize well to new data. This performance degradation between the training and test phases highlights the importance of model validation and suggests the need for regularization techniques or hyper parameter tuning.

Table 3. Xgboost Regression Weekly Pay Train And Test Performance Metrics

The XGBoost model shows exceptional performance on the training data with an R² value of 0.9997, indicating that the model explains all the variation in weekly wages. The training metrics demonstrate excellent accuracy with $2.19 RMSE and $1.15 MAE, indicating that the predictions deviate minimally from the true values. The maximum error of $5.00 and the mean absolute error of $0.0012 further confirm the accuracy of the model during training. These metrics indicate that the algorithm has successfully learned the underlying relationships between employee characteristics and compensation. However, the test performance exhibits significant degradation, highlighting potential overfitting concerns. The R² drops significantly to 0.9011, however, indicating still strong predictive ability, explaining approximately 90% of the salary variation in the unobserved data. The dramatic increase in RMSE to $49.14 and MAE to $44.17 indicates significantly larger prediction errors on the new data. The maximum error increases to $80.00, indicating that the model struggles with some salary predictions, especially at extreme values.The explained variance score (EVS) shows a less dramatic degradation from 0.9997 to 0.9694, indicating that the model maintains a reasonable ability to capture the overall variance structure in the test data. The mean square log error (MSLE) is relatively low at 0.0026 for the test, indicating that the model adequately handles the logarithmic scale of salary predictions.The performance gap between the training and testing phases indicates that the model has memorized training-specific patterns rather than learning general relationships. The average absolute error of $45.00 on the test data indicates that half of all predictions deviate by less than this amount, which represents a 5% error of about $863 for average salaries. Although the model shows promise with 90% variance explanation in new data, the significant performance degradation indicates that regularization techniques, cross-validation, or hyper parameter tuning are needed to improve generalization and reduce overfitting for practical use in salary forecasting applications.

FIGURE 5.XGBoost Regression Weekly Pay Training

The XGBoost model demonstrates superior training performance compared to Gradient Boosting, with data points forming a nearly perfect linear relationship in the reference series. This algorithm successfully predicts weekly wages across the entire range with minimal deviation, suggesting superior learning of the underlying compensation models. The model’s ability to accurately predict entry-level and senior-level salaries indicates robust feature utilization and effective handling of relationships between hours worked, skill levels, and overtime compensation. Tight clustering demonstrates XGBoost’s ensemble learning capabilities in capturing complex interactions between predictor variables.

FIGURE 6.XGBoost Regression WeeklyPay Testing

The XGBoost test results show significantly better generalization compared to the gradient boosting model, although some performance degradation is still evident. Predictions maintain reasonable accuracy across most salary ranges, with particularly good performance in the $800-1100 range, where most employees fall. However, the model still struggles with extreme values, showing some prediction errors for both low- and high-paid employees. The overall method indicates better regularization and generalization capabilities than gradient boosting, but indicates that predicting compensation at salary peaks remains a challenge for both models, perhaps due to the limited training examples in these ranges.

Conclusion

The development and implementation of team pay analytics and resource optimization platforms represent a transformative approach to modern workforce management, especially in complex operational environments such as aviation, utilities, and field services. This research demonstrates that sophisticated machine learning techniques, specifically gradient boosting and XGBoost algorithms, can effectively model compensation structures and predict weekly pay with high accuracy. The strong correlations identified between employee characteristics - working hours, skill levels, overtime hours - and compensation outcomes confirm the effectiveness of competency-based pay systems while providing operational insights for organizational decision-making. The comparative analysis between gradient boosting and XGBoost reveals important considerations for practical implementation. While both models achieved nearly perfect training performance with R² values of 0.9997, the testing phase revealed significant overfitting challenges that are characteristic of ensemble learning applications. XGBoost’s superior generalization capabilities, demonstrated by excellent error control and built-in regularization algorithms, position it as the preferred algorithm for production environments. However, the performance degradation observed in both models emphasizes the importance of implementing proper validation techniques, cross-validation procedures, and hyper parameter optimization strategies. The strategic value of these platforms extends beyond mere compensation prediction to include comprehensive workforce optimization, regulatory compliance, and employee satisfaction. Despite the overriding concerns, the ability to achieve 90% prediction accuracy on unobserved data demonstrates the practical viability of these systems for HR analytics and strategic workforce planning. Future implementations should prioritize advanced regularization techniques, feature selection methods, and ensemble approaches to improve generalization while maintaining predictive power. As organizations begin to navigate increasingly complex operational demands and regulatory requirements, group payroll analytics platforms will become essential tools for achieving operational efficiency, cost control, and employee satisfaction in the digital transformation era.

REFERENCE

1. Balwani, Siddharth SiddharthVashdev. "Operational efficiency through resource planning optimization and work process improvement." PhD diss., Massachusetts Institute of Technology, 2012.
2. Lazear, Edward P., and Kathryn L. Shaw. "Personnel economics: The economist's view of human resources." Journal of economic perspectives 21, no. 4 (2007): 91-114.
3. Hamilton, Barton H., Jack A. Nickerson, and Hideo Owan. "Team incentives and worker heterogeneity: An empirical analysis of the impact of teams on productivity and participation." Journal of political Economy 111, no. 3 (2003): 465-497.
4. Gopalakrishnan, Balaji, and Ellis L. Johnson. "Airline crew scheduling: State-of-the-art." Annals of Operations Research 140, no. 1 (2005): 305-337.
5. PK Kanumarlapudi. (2023) Strategic Assessment of Data Mesh Implementation in the Pharma Sector: An Edas-Based Decision-Making Approach. SOJ Mater Sci Eng 9(3): 1-9. DOI: 10.15226/2473-3032/9/3/00183
6. Adzic, Gojko, and Robert Chatley. "Serverless computing: economic and architectural impact." In Proceedings of the 2017 11th joint meeting on foundations of software engineering, pp. 884-889. 2017.
7. Davenport, Thomas H. "Competing on analytics." Harvard business review 84, no. 1 (2006): 98.
8. Bonomi, Flavio, Rodolfo Milito, Preethi Natarajan, and Jiang Zhu. "Fog computing: A platform for internet of things and analytics." In Big data and internet of things: A roadmap for smart environments, pp. 169-186. Cham: Springer International Publishing, 2014.
9. Peram, S. R. (2024). Automated Label Detection and Recommendation System Using Deep Convolution Neural Networks and SPSS-Based Evaluation. International Journal of Computer Science and Data Engineering, 1(2), 258. https://doi.org/10.55124/csdb.v1i1.258
10. Enemosah, Aliyu. "REDUCING DOWNTIME WITH CLOUD-NATIVE PREDICTIVE ANALYTICS FOR OFFSHORE EQUIPMENT MONITORING IN OIL AND GAS."
11. Wei, Chun-Chin, Chen-Fu Chien, and Mao-Jiun J. Wang. "An AHP-based approach to ERP system selection." International journal of production economics 96, no. 1 (2005): 47-62.
12. Joseph, Taiwo. "Optimizing Operational Expenditure (OPEX) in Offshore Support Vessel (OSV) Operations: A Benchmarking Study Against International Best Practices." (2025).
13. Pamulaparti Venkata, S. "Optimizing Resource Allocation For Value-Based Care (VBC) Implementation: A Multifaceted Approach To Mitigate Staffing And Technological Impediments Towards Delivering High-Quality, Cost-Effective Healthcare." Australian Journal of Machine Learning Research & Applications 3, no. 2 (2023): 304-330.
14. Wang, Wanyuan, Zhanpeng He, Peng Shi, Weiwei Wu, Yichuan Jiang, Bo An, Zhifeng Hao, and Bing Chen. "Strategic social team crowdsourcing: Forming a team of truthful workers for crowdsourcing in social networks." IEEE Transactions on Mobile Computing 18, no. 6 (2018): 1419-1432.
15. Goyal, Abhishek. "Optimising cloud-based CI/CD pipelines: Techniques for rapid software deployment." Int J Eng Res 11, no. 11 (2024): 896-904.
16. Atluri, Anusha, and Vijay Reddy. "Total Rewards Transformation: Exploring Oracle HCM's Next-Level Compensation Modules." International Journal of Emerging Research in Engineering and Technology 4, no. 1 (2023): 45-53.
17. Seibert, Scott E., Gang Wang, and Stephen H. Courtright. "Antecedents and consequences of psychological and team empowerment in organizations: a meta-analytic review." Journal of applied psychology 96, no. 5 (2011): 981.
18. Fan, Chao, Diwei Liu, Rui Huang, Zhigang Chen, and Lei Deng. "PredRSA: a gradient boosted regression trees approach for predicting protein solvent accessibility." Bmc Bioinformatics 17, no. Suppl 1 (2016): S8.
19. Kakulavaram, S. R. (2024). “Intelligent Healthcare Decisions Leveraging WASPAS for Transparent AI Applications” Journal of Business Intelligence and DataAnalytics, vol. 1 no. 1, pp. 1–7. doi: https://dx.doi.org/10.55124/csdb.v1i1.261
20. Liao, Shengli, Zhanwei Liu, Benxi Liu, Chuntian Cheng, Xinfeng Jin, and Zhipeng Zhao. "Multistep-ahead daily inflow forecasting using the ERA-Interim reanalysis data set based on gradient-boosting regression trees." Hydrology and Earth System Sciences 24, no. 5 (2020): 2343-2363.
21. Peram, S. R (2023). Advanced Network Traffic Visualization and Anomaly Detection Using PCA-MDS Integration and Histogram Gradient Boosting Regression. Journal of Artificial Intelligence and Machine Learning, 1(3), 281. https://doi.org/10.55124/jaim.v1i3.281
22. Moisen, Gretchen G., Elizabeth A. Freeman, Jock A. Blackard, Tracey S. Frescino, Niklaus E. Zimmermann, and Thomas C. Edwards Jr. "Predicting tree species presence and basal area in Utah: a comparison of stochastic gradient boosting, generalized additive models, and tree-based methods." Ecological modelling 199, no. 2 (2006): 176-187.
23. Feng, Ji, Yang Yu, and Zhi-Hua Zhou. "Multi-layered gradient boosting decision trees." Advances in neural information processing systems 31 (2018).
24. Deng, Lei, Wenyi Yang, and Hui Liu. "PredPRBA: prediction of protein-RNA binding affinity using gradient boosted regression trees." Frontiers in genetics 10 (2019): 637.
25. Li, ZhuoXuan, XinLi Shi, JinDe Cao, XuDong Wang, and Wei Huang. "CPSO-XGBoost segmented regression model for asphalt pavement deflection basin area prediction." Science China Technological Sciences 65, no. 7 (2022): 1470-1481.
26. PK Kanumarlapudi. “Improving Data Market Implementation Using Gray Relational Analysis in Decentralized Environments” Journal of Artificial intelligence and Machine Learning., 2024, vol. 2, no. 1, pp. 1–7. doi: https://dx.doi.org/10.55124/jaim.v2i1.271
27. Li, ZhuoXuan, XinLi Shi, JinDe Cao, XuDong Wang, and Wei Huang. "CPSO-XGBoost segmented regression model for asphalt pavement deflection basin area prediction." Science China Technological Sciences 65, no. 7 (2022): 1470-1481.
28. Kim, Yongjun, and Yungcheol Byun. "Predicting solar power generation from direction and tilt using machine learning xgboost regression." In Journal of Physics: Conference Series, vol. 2261, no. 1, p. 012003. IOP Publishing, 2022.
29. Truong, Gia Toai, Kyoung-Kyu Choi, Tri-Hai Nguyen, and Chang-Soo Kim. "Prediction of shear strength of RC deep beams using XGBoost regression with Bayesian optimization." European Journal of Environmental and Civil Engineering 27, no. 14 (2023): 4046-4066.
30. Pramanik, Paritosh, Rabin K. Jana, and Indranil Ghosh. "AI readiness enablers in developed and developing economies: Findings from the XGBoost regression and explainable AI framework." Technological Forecasting and Social Change 205 (2024): 123482.