The identification of talent in sports is a matter of much scientific and practical importance, and, ideally, to efficiently allocate the scarce resources of the training process, excellent athletes should be selected [1,2]. The specificity of anthropometric, physical and technical qualities that determine the effectiveness of athletes in volleyball is an additional factor that contributes to the uniqueness of player selection patterns [3,4]. The common solutions to this problem are often extensive groupings of tests used in talent identification measuring jumping, agility, endurance and skill performance [5,6]. Still, such approaches can be less efficient, too time-wasting and non-prediction-based in terms of the performance results [7,8].

The novel developments in the realm of artificial intelligence (AI) and machine learning (ML) have opened up some prospects in terms of talent identification in sport. Machine learning algorithms will improve the processing of complicated data and simplify the decision-making process when choosing the most optimal trait that should be used by coaches and talent triers [9,10]. The ML applications in the areas such as basketball, football, table tennis and boxing have already shown the potential of using data-driven techniques to discover important predictors of performance and to optimally design training strategies [11-14]. The volleyball research field is also starting to move in this direction, as recent works focus on how to combine physical and tactical variables in more coherent models of assessing a player [15-17].

Although there is an increased awareness of the role of ML in sports analytics, few researchers have specifically addressed the role of ML in helping select the best fitness test to identify talented volleyball players. Current common methods include a general battery of tests or focus on individual characteristics without checking their combined predictive validity of a higher level of performance [6,18,19]. This loophole brings about inefficiencies in the process of assessing the athlete and the possibilities of missing vital performance indicators which bring a difference between the stars and the rest. It is therefore of utmost significance to develop a comprehensive machine learning-driven study that will determine the most informative sub-set of physical fitness tests pertinent to volleyball and minimize redundancy within physical fitness tests, yet, maximize predictive variance and accuracy.

This paper will be devoted to the discussion of three key aims:

• To gather and assess a broad range of data representing anthropometrical, physical and skill-based tests of about 200 volleyball players of various age groups

• To highlight and compare the effectiveness of machine learning algorithms (Random Forest, Support Vector Machine and XGBoost) to identify which tests apply best in slotting the seemingly unlimited number of players who want to play volleyball at the elite level

• To design and confirm a short, effective test battery that could enable coaches and sports organization to use fewer resources to identify talent with reasonable accuracy

There are three major contributions of this study. First, it synthesizes sports science and advanced ML to create an evidence-based framework of volleyball talent scouting, which solves the shortcomings of the original techniques [16,17]. Second, it substantiates the predictive validity of certain aptitude and ability tests beyond evaluation criteria (ROC/AUC), that attains both accuracy and generalizability. Lastly, it makes practical recommendations to practitioners so that the streamlined, cost-effective and scientifically validated testing protocol could be implemented.

The uniqueness of this work is that it uses machine learning to optimise physical fitness test selection in volleyball, which is an area where this form of integration is rather inexhaustive. As opposed to other researchers who either evaluate each physical quality individually [5,15,20,21] or use coach ratings to create a composite score of sport-related qualities [22-24], the current study deploys ML algorithms to determine the fewest number of tests to achieve the greatest predictive value. The examination of this way fulfils the research on talent identification in terms of methodological rigor and the lack of a connection between theory and practical actions. In addition, the framework utilization of both cross-validation and multiple ML models would add to the applicability of the work to various populations of athletes, furthering the sphere of sports performance facilitated by technology [9,25].

Identifying talent has long been viewed as a fundamental cornerstone to competition-based athletics, seeking out those with the greatest ability to become competitive in the long-term. The initial models in football, basketball and volleyball focused on anthropometric, physical and technical tests [1,2,22]. In volleyball, it all depends on body size in combination with both motor coordination and tactical understanding [16,17]. It has been found that the ability to advance to elite level performances requires both physical and perceptual-cognitive properties [3]. Also, the multidimensional assessment aspect is prevalent in cross-sport studies, e.g., in basketball [12,18] and rugby [8].

Specific quality research studies on physical fitness always associate certain physical fitness attributes with success in volleyball. To cite an example, the power of the upper and lower body, as determined by vertical jumps and strength tests, is a good indicator of the elite selection [15]. Likewise, agility, coordination and explosive movements differ with respect to players of different competitive levels [5]. Similar findings were found in the work of Oliveira et al. [4] regarding the variation of change of direction deficits with age in female volleyball players and primary research by Rubajczyk and Rokita [7], who stressed on the importance of relative age in talent selections. Along with these physical factors, neuromuscular training is also useful in the development of young people [26]. In addition, it is imperative to measure technical skills as well; the systematic reviews indicate that passing, setting and spiking accuracy should also be evaluated to conduct a comprehensive talent identification [6,19].

Talent identification goes beyond selection trends in a physical capacity The key predictor of elite progression in young players therefore is their cognitive and perceptual-cognitive development [3,27]. The more general sociocultural literature in sports and education shows how gender, opportunity and social norms determine access to training and competition [28-30]. These views have echoes in cultural approaches to modernity [31], ethics [32] and the Category of regenerating the city [33-35]. Although the non-sport sources of some works are not limited to sport [36-38], they aim to define the wider frameworks and decision-making motives sports professionals involved in athlete development and resource distribution.

Artificial intelligence and machine learning are factors that have recently boosted sports analytics because of their ability to provide data-driven commonalities in performance analysis and in the development of athletes [9,25]. Examples of them include an optimization of serve strategies in table tennis [11], intelligent boxing gloves to classify the moves [13,23] and prediction of the market value in football [10]. In sports such as volleyball, ML is being used more and more to tune training programs [14], to improve tactical decision-making [8] and to determine the most essential performance determinants [17]. Such progress indicates that incorporation of ML into the volleyball talent identification could help to step beyond the use of descriptive statistics towards predictive and evidence-based tools.

There are still some issues in the development of streamlined and validated batteries of tests. The use of large numbers of tests in the current volleyball research has a tendency of failing to adequately validate the predictive significance of the testing [5,6]. With multidimensional models, though some exist [16], these are never simplified into efficient and coach-friendly protocols. Besides, previous methods often lack contextual biases including relative age [7] or age variations in development [4]. Including machine learning has the potential to fill these gaps as it would reduce redundancy and involve the most significant performance indicators.

Even though the given study is volleyball-specific, the experiences of other domains testify to the significance of the holistic and data-based experiences. As an example, use of multidimensional assessment models to inform decision-making in clinical and mental studies has been demonstrated [39-41]. Equally, cultural examination of performance and identity [42,43] points to the meeting of physical capability and factors of social and psychological nature. These cross-disciplinary reflections remind us that becoming a talent is not an exclusive process of physiological detection but is also contextualized through more diffuse regimes of assessment and acknowledgement.

To systematically determine how various physical fitness measures predict volleyball talent, this research study utilised a systematic methodology that is informed both by sports science and machine learning. The methodology was preset to be reproducible, rigorous and practically applicable, as the study passed sampling of the subjects, measurement of the parameters and mathematical modelling, to the creation of an optimized testing system.

To give a comprehensive picture of the methodological steps, Figure 1 summarizes the suggested machine learning-grounded approach to the selection of the best physical fitness tests in volleyball with consideration of individuals with talent in sports. The flowchart presents the sequential process of data set collection to the model evaluation towards a generation of a reduced and lean test battery.

Study Design and Participants

The current research involved a cross-sectional study with an anthropometric, physiological and skill-based measurement and modelling of machine learning. Around ~200 volleyball players (male and female) were recruited in regional academies designate, youth development centres and professional clubs. They were stratified by three different age groups (U16, U20, Senior) because of developmental differences [4].

Figure 1: Proposed machine learning-driven framework for volleyball talent identification

Inclusion Criteria

• At least 3 years of systematic volleyball training

• Compete at an official level in the past 12 months

• Lack of injuries which limit performance

Exclusion Criteria

• Fragile test information

• Recent rehabilitation against musculoskeletal injury

The institutional review board granted the ethics approval and informed consent (in case of minors) was obtained.

Data Collection

The multidimensional qualities of volleyball players were taken into consideration when the process of data collection was designed. This was done by incorporating anthropometric, physiological and skill-based parameters which helped base a rich source of data that will depict the physical, technical and functional requirements of volleyball performance at high levels.

Anthropometric Variables

• Height (cm), Body mass (kg), Body mass index (BMI)

• Arm span (cm), Sitting height (cm)

• Body fat percentage, body lean mass by bioelectrical impedance

Physical Fitness Tests

• Vertical Jump (cm): Countermovement jump with arms (CMJA) and without arms (CMJ)

• Agility T-Test (s): time spent in passing a standard agility course

• Handgrip Strength (kg): The maximum amount of the force grasped by a dynamometer

• VO₂max (ml·kg⁻¹·min⁻¹): Estimated from a Yo-Yo intermittent recovery test

• Sprint Speed (20 m, 10 m in s)

• Spike Speed (km/h): Obtained with the use of a radar gun when doing maximal spike attempts

Skill-Based Measures

• Service accuracy (%): Number of successful serves targeting zones

• Passing efficacy (%): Based on coach-rated accuracy scale

• Spiking accuracy (%): Successful spikes under controlled conditions

Test Battery Construction

A comprehensive test battery of ≈15-20 variables was constructed combining anthropometric, physical and skill tests. In order to standardize comparisons for all these variables, they were all z-normalized:

where xi = raw test score, μ = mean, σ = standard deviation.

Machine Learning Modelling

Three ML algorithms were used:

• Random Forest (RF)

• Support Vector Machine (SVM)

• Extreme Gradient Boosting (XGBoost)

Random Forest (RF)

An ensemble of decision trees, where prediction is based on majority voting:

where ht(x) is the prediction from the t-th decision tree and T is the number of trees.

Feature importance was computed via Gini impurity reduction:

where pk is the probability of class k.

Support Vector Machine (SVM)

The classifier attempts to find a hyperplane:

subject to:

where w = weight vector, b = bias, = class labels. The kernel trick (Radial Basis Function, RBF) was used:

Extreme Gradient Boosting (XGBoost)

The regularized objective function that minimizes:

with:

where l = loss function, Ω = regularization term, T = number of leaves, w = leaf weights, γ, λ = penalty terms.

Model Training and Validation

• Data split: 70% training and the rest 30% for testing

• Cross-validation: 10-fold stratified CV to reduce bias

• Hyperparameter tuning: Grid search on parameters such as tree depth, learning rate and kernel parameters

Performance measured using:

• Accuracy (ACC)

• Precision (P), Recall (R), F1-score

• ROC Curve & AUC

Accuracy (ACC)

Precision (P), Recall (R), F1-score

 ​
 

ROC Curve & AUC

where TPR = true positive rate, FPR = false positive rate.

Feature Selection and Optimal Test Battery

RF and XGBoost variable importance were obtained and SVM feature weights were extracted. An aggregation of rank measure was employed:

Score(v) = final score of variable v.

The minimal optimal test battery was the lowest number of tests that attained the predictive performance of the full model (=95).

Practical Implementation

Finally, this shorter test battery was discussed with expert coaches in terms of practical feasibility (time, cost, equipment requirements). It was recommended that a balance should be provided between predictive accuracy and field applicability.

This paper reports the findings of the research in a clear structure where descriptive statistics are displayed to give an overview of the characteristics of players in the various age groups. The efficacy of machine learning models and the significance of each test are also analysed to indicate the most promising qualities that are used in volleyball talent detection.

Descriptive Statistics

Table 1 presents the descriptive statistics to the sample of volleyball players indication the anthropometric and physical performance features according to three age groups (U16, U20, Senior).

Interpretation

The critical development differences were found in almost all the variables, which concurred with the previously established results that maturation improves anthropometric and neuromuscular qualities in volleyball players [4,15].

Machine Learning Model Performance

Table 2 shows the accuracy of prediction of the three machine learning algorithms. Figure 2 shows how Random Forest, SVM and XGBoost perform in terms of classification compared with several metrics, once again showing the dominance of XGBoost.

Interpretation: 

XGBoost performed best overall (AUC = 0.95), as it can deal with both nonlinear dependencies and heterogeneous data [9,10].

Figure 3 shows ROC curves in the Random Forest model, SVM and XGBoost. Since the area under the curve of XGBoost was the largest, it was proven that it was the best at discriminating.

Feature Importance and Optimal Test Battery

The top predictors as obtained by the feature importance ranking attributed to Random Forest and XGBoost with SVM weights included.

Figure 4 shows the relative amounts of the most influential variables in all types of machine learning. Spike speed and vertical jump came on top of the list of predictors followed by agility and handgrip strength.

Interpretation: 

Spike speed, vertical jump, agility and handgrip strength were the most powerful predictors of volleyball talent, which agree with the biomechanical and neuromuscular requirements of the sport [16,17].

Reduced Test Battery

The optimal reduced test battery was considered the minimal associated test-battery to attain an accuracy equivalent to that of the full model (i.e., 95% or more).

REDUCED TEST BATTERY (5 tests):

• Spike Speed

• Vertical Jump

• Agility T-test

• Handgrip Strength

• Height

This combination preserved an AUC of 0.93, which is almost lower than the full battery (AUC = 0.95).

Table 1: Descriptive statistics of key performance variables (mean ± SD)

*Significant at p < 0.05.

Table 2: Performance of ML models in predicting high-level volleyball performance

Table 3: Ranked importance of predictors (aggregated across models)

Figure 2: Comparison of ML models’ performance in predicting high-level volleyball performance

Figure 3: ROC curves for ML models in predicting volleyball performance

The findings indicate that machine learning can significantly maximize the talent identification processes in volleyball. Traditional methods usually utilize comprehensive test batteries that are resource-consuming [5,6]. In comparison, the ML-based scheme here worked effectively in suggesting a parsimonious scheme of tests that maximized predictive efficiency without increasing redundancy. The fact that the predictors spike speed and vertical jump became dominant emphasizes the previous biomechanical findings, which indicate how explosive power can precondition the high performance in volleyball [7,15]. Handgrip strength also served to discriminate among higher-level athletes, with agility provided a secondary contribution, highlighting the significance of reactive movement and upper-limb strength in execution of the technical aspects of sport.

These are consistent with multidimensional views of talent identification [12,16], that elite potential is a product of both neuromuscular and anthropometric qualities. Furthermore, the employment of cross-validation and ROC analysis allowed checking the validity of the findings and further advanced the previous studies where the results were based more on linear relationships [1,8].

In practice, the shortened test battery helps a coach or a federation to resolve the central dilemma involving scientific accuracy on one hand and the logistical practicality on the other. The proposed framework does not necessitate much time, resources or equipment to be used and still direct talent selection in an evidence-based manner. This time-saving functionality reflects a similar pattern across the sports industry that have embraced AI and ML to automate performance assessments [11,13,14].

Lastly, the findings are applicable to the wider discussions about the role of AI in sports analytics where such tools are already driving decision-making and predictions [9,25]. The method has relevance not only to the sport of volleyball but to talent identification based on machine intelligence in general.

Since physical fitness test batteries possess the potential of improving volleyball talent identification, this study showed the usefulness of a machine learning-based framework to optimize physical fitness test batteries. Combining anthropometric, physical and skills data in a group of mixed players, it was found that maximum spike speed, vertical jump, agility, handgrip strength and height were the strongest predictors of high-level performances. Most importantly, the shorter test battery generated as much predictability as the entire test battery in a way that demonstrates that artificial intelligence can have the potential to simplify the process of identifying talent by streamlining those processes without impairing scientific standards.

The studies contribute to existing research because of shifting the focus off canonical linear models and toward the robust machine learning algorithms, such as Random Forest, Support Vector Machines and XGBoost. The strong performance of XGBoost supported the relevance of the models that are able to consider nonlinear interactions and intricate dependencies between performance attributes. These observations serve not only to re-establish the importance of explosive power, agility and strength, in volleyball performance but also offer an empirical blueprint that can be delivered to coaches, scouts and federations immediately.

In practical regards, the study provides a practically realizable and evidence-based recourse that could be applied by sports organisations faced by limited sources of time and resources. The flexibility of the framework is that it can be used effectively in conducting large-scale screening, early talent identification as well as in planning over the long-term development of players involved by focusing on a parsimonious test battery. The methodological approach can be broadened to other team sports, in which the talent identification process is an expensive, but essential process.

Finally, the study points out the paradigm shift that machine learning has had on the field of sports science. By filling the gap between innovative methods of computations and practical talent evaluation, the work has theoretical significance to the body of knowledge and practical applications in practice. Future study should also apply the idea to various groups, pay attention to psychological and tactical aspects of talent and assess the results in the long-term perspective to demonstrate the optimal validity of the minimized test array.

1. Sarmento, H. et al. “Talent identification and development in male football: A systematic review.” Sports Medicine, vol. 48, no. 4, 2018, pp. 907-931.

2. Williams, A.M. et al. “Talent identification and development in soccer since the millennium.” Science and Football, 2023, pp. 3-14.

3. De Waelle, S. Inside the Mind of the Young Volleyball Player: The Development of Cognitive and Perceptual-Cognitive Function. Ghent University, 2021. Doctoral dissertation.

4. Oliveira, J.P. et al. “Characterization of physical performance and change of direction deficit across age groups in young female volleyball players.” BMC Sports Science, Medicine and Rehabilitation, vol. 17, no. 1, 2025, p. 209.

5. Farley, J.B. et al. “The relationship between physical fitness qualities and sport-specific technical skills in female, team-based ball players: A systematic review.” Sports Medicine-Open, vol. 6, no. 1, 2020, p. 18.

6. Koopmann, T. et al. “Assessing technical skills in talented youth athletes: A systematic review.” Sports Medicine, vol. 50, no. 9, 2020, pp. 1593-1611.

7. Rubajczyk, K. and A. Rokita. “The relative age effect and talent identification factors in youth volleyball in Poland.” Frontiers in Psychology, vol. 11, 2020, p. 1445.

8. Pino-Ortega, J. et al. “Training design, performance analysis and talent identification—A systematic review about the most relevant variables through the principal component analysis in soccer, basketball and rugby.” International Journal of Environmental Research and Public Health, vol. 18, no. 5, 2021, p. 2642.

9. Pietraszewski, P. et al. “The role of artificial intelligence in sports analytics: A systematic review and meta-analysis of performance trends.” Applied Sciences, vol. 15, no. 13, 2025, p. 7254.

10. Tamim, A. et al. “Machine learning-driven market value prediction for European football players.” Journal of Computational Mathematics and Data Science, 2025, p. 100118.

11. Chang, C.W. et al. “Deep learning-driven decision tree ensembles for table tennis: Analyzing serve strategies and first-three-stroke outcomes.” Computer Science and Information Systems, 2025, pp. 40-40.

12. Joseph, J. et al. “A comparison of multidimensional qualities discriminant of selection in elite adolescent Australian basketball athletes.” PLOS One, vol. 16, no. 8, 2021.

13. Cizmic, D. et al. “Smart boxing glove ‘RD α’: IMU combined with force sensor for highly accurate technique and target recognition using machine learning.” Applied Sciences, vol. 13, no. 16, 2023, p. 9073.

14. Wan, L. “Design and adjustment of optimizing athletes’ training programs using machine learning algorithms.” Journal of Electrical Systems, vol. 20, no. 6s, 2024, pp. 2014-2024.

15. Tsoukos, A. et al. “Upper and lower body power are strong predictors for selection of male junior national volleyball team players.” The Journal of Strength & Conditioning Research, vol. 33, no. 10, 2019, pp. 2760-2767.

16. Sgrò, F. et al. “A multidimensional approach to talent identification in youth volleyball through declarative tactical knowledge and functional fitness.” Journal of Functional Morphology and Kinesiology, vol. 9, no. 1, 2024, p. 29.

17. Hajilou, B. and M. Anbarian. “Investigate the most important characteristics of elite volleyball players for sports talent identification.” Journal of Advanced Sport Technology, vol. 7, no. 4, 2023, pp. 1-10.

18. Sodaitis, B. Relation between Anthropometric, Physical, Technical Testing and Game-Related Statistics in Youth Basketball Players. Lietuvos sporto universitetas, 2020. Doctoral dissertation.

19. Mohammed Hammood, Y. et al. “The effect of a proposed training method using play exercises to develop specific agility and skill performance in football | El efecto de un método de entrenamiento propuesto que utiliza ejercicios de juego para desarrollar la agilidad específica y el re.” Retos, vol. 63, 2025, pp. 719-728. https://doi.org/10.47197/retos.v63.111095.

20. Khalaf, Y.A. et al. “The effect of sports job burnout on the performance of workers in student activities departments in Iraqi universities | El efecto del agotamiento laboral deportivo en el rendimiento de los trabajadores de los departamentos de actividades estudiantiles de.” Retos, vol. 66, 2025, pp. 86-95. https://doi.org/10.47197/retos.v66.113271.

21. Omar, A.F. et al. “The impact of cognitive training on field intelligence growth and some composite skills of advanced football players | El impacto del entrenamiento cognitivo en el crecimiento de la inteligencia de campo y algunas habilidades compuestas de los jugadores d.” Retos, vol. 66, 2025, pp. 46-58. https://doi.org/10.47197/retos.v66.113234.

22. Larkin, P. and D. O’Connor. “Talent identification and recruitment in youth soccer: Recruiter’s perceptions of the key attributes for player recruitment.” PLOS One, vol. 12, no. 4, 2017.

23. Ali, O.A. et al. “The effect of a proposed training curriculum to develop some special physical abilities and the accuracy of the movement scoring skill for the Ramadi football club players | El efecto de un plan de estudios de entrenamiento propuesto para desarrollar algu.” Retos, vol. 61, 2024, pp. 193-200. https://doi.org/10.47197/retos.v61.107271.

24. Hammood, Y.M. et al. “Measuring the aggressive behavior of the teams in the Iraqi Premier League in football and its relation to the results and ranking of the league for the 2022-2023 season.” Sportske Nauke i Zdravlje, vol. 14, no. 2, 2024, pp. 127-134. https://doi.org/10.7251/SSH2402127H.

25. Liu, H. “Technology-based sports performance enhancement: A conference review.” Intelligent Sports and Health, vol. 1, no. 2, 2025, pp. 98-102.

26. Zemková, E. and D. Hamar. “Sport-specific assessment of the effectiveness of neuromuscular training in young athletes.” Frontiers in Physiology, vol. 9, 2018, p. 264.

27. AbdulJabbar, M.A. et al. “El efecto del entrenamiento mental (visualización mental y enfoque de atención) en la precisión y velocidad del rendimiento ofensivo en jugadores de esgrima.” Retos, vol. 70, 2025, pp. 1097-1113. https://doi.org/10.47197/retos.v70.117026

28. Smith, S.W. The Segregation of Girls in Mathematics. Sheffield Hallam University (United Kingdom), 1986. Doctoral dissertation.

29. May, L. Masculinity & Morality. Cornell University Press, 1998.

30. Berwick, B.E. and C. Oppenheimer. “Marriage pregnancy and the right to go to school.” Texas Law Review, vol. 50, 1971, p. 1196.

31. Naremore, J. and P.M. Brantlinger, editors. Modernity and Mass Culture. Indiana University Press, 1991.

32. Hult, R.A. Agnosticity Volume 2: An Agnostic View of Bothersome Mormon Doctrine. Vol. 2, Trafford Publishing, 2007.

33. Warren, A. Urban Regeneration in the 1990s: A Review of the City Challenge Initiative. University of London, University College London (United Kingdom), 1993. Doctoral dissertation.

34. Sagalyn, L.B. Times Square Roulette: Remaking the City Icon. MIT Press, 2003.

35. Hussein Fayyad, F. et al. “Construir y legalizar una prueba para medir el nivel de agilidad futbolística de los jugadores jóvenes.” Retos, vol. 68, 2025, pp. 1578-1590. https:// doi. org/10.47197/retos.v68.116368

36. Garner, W.R. The Subjective Prestige Factor in the Diplomatic Milieu of the Chaco War: A Case Study in Systems Analysis. Tulane University, 1963.

37. Satloff, R. The Battle of Ideas in the War on Terror. Washington Institute for Near East Policy, 2004.

38. Wang, F.Y. The Art of State Persuasion: China's Strategic Use of Media in Interstate Disputes. Oxford University Press, 2024.

39. Roberts, E.M. Mental Health Clinical Services; A Study of the Children between 6 and 12 Years of Age Examined by Mental Health Clinics in Vancouver from 1945 to 1947 Inclusive. University of British Columbia, 1949. Doctoral dissertation.

40. Kelly, George et al. The Handbook of Clinic Practice. 6th Edn., New York, Fort Hays State University, ISBN-13: 9781462547043, http://dx.doi.org/10.58809/saar9809.

41. Friedenberg, E.S. “Jaws III: The impropriety of shark-repellent amendments as a takeover defense.” Delaware Journal of Corporate Law, vol. 7, 1982, p. 32.

42. Tornatore, G. Becoming a “Jazzperson”: Moving beyond Sounds and Tones. Teachers College, Columbia University, 2013. Doctoral dissertation.

43. Phoutchanthavongsa, S. Lesbian, Gay, Bisexual and Transgender (LGBT) Disclosure among the Global Fortune 250 Companies - An Exploration. Auckland University of Technology, 2019. Doctoral dissertation.