ArticlePublisher preview available

Exploiting sensor data in professional road cycling: personalized data-driven approach for frequent fitness monitoring

Authors:
Arie-Willem de Leeuw at The Hague University of Applied Sciences
Arie-Willem de Leeuw
Mathieu Heijboer at Team Jumbo-Visma
Mathieu Heijboer
  • Team Jumbo-Visma
Tim Verdonck at KU Leuven
Tim Verdonck
Arno Knobbe at Leiden University
Arno Knobbe
Show all 5 authorsHide
Read publisher preview
Download citation
To read the full-text of this research, you can request a copy directly from the authors.

Abstract and Figures

We present a personalized approach for frequent fitness monitoring in road cycling solely relying on sensor data collected during bike rides and without the need for maximal effort tests. We use competition and training data of three world-class cyclists of Team Jumbo–Visma to construct personalised heart rate models that relate the heart rate during exercise to the pedal power signal. Our model captures the non-trivial dependency between exertion and corresponding response of the heart rate, which we show can be effectively estimated by an exponential kernel. To construct the daily heart rate models that are required for day-to-day fitness estimation, we aggregate all sessions in the previous week and apply sampling. On average, the explained variance of our models is 0.86, which we demonstrate is more than twice as large as for models that ignore the temporal integration involved in the heart’s response to exercise. We show that the fitness of a cyclist can be monitored by tracking developments of parameters of our heart rate models. In particular, we monitor the decay constant of the kernel involved, and also analytically determine virtual aerobic and anaerobic thresholds. We demonstrate that our findings for the virtual anaerobic threshold on average agree with the results of exercise tests. We believe this work is an important step forward in performance optimization by opening up avenues for switching to adaptive training programs that take into account the current physiological state of an athlete.
Cumulative probability density function for the convolved pedal power if we combine all training sessions of the cyclists. As an example, we considered a recuperation time τ=40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau = 40$$\end{document} s and considered a convolution window of 120 s. We observe that high values for the convolved pedal power are scarce for all three cyclists
Cumulative probability density function for the convolved pedal power if we combine all training sessions of the cyclists. As an example, we considered a recuperation time τ=40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau = 40$$\end{document} s and considered a convolution window of 120 s. We observe that high values for the convolved pedal power are scarce for all three cyclists
… 
Illustration of the deflection points PLT1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{LT1}}$$\end{document} and PLT2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{LT2}}$$\end{document} in an S-curve for the relationship between the heart rate and the convolved pedal power. The deflection points are defined as the points on the S-curve that have the largest perpendicular distance to the straight lines drawn between Pmin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{min}}$$\end{document} and Plin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{lin}}$$\end{document} or Plin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{lin}}$$\end{document} and Pmax\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{max}}$$\end{document}, respectively
Illustration of the deflection points PLT1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{LT1}}$$\end{document} and PLT2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{LT2}}$$\end{document} in an S-curve for the relationship between the heart rate and the convolved pedal power. The deflection points are defined as the points on the S-curve that have the largest perpendicular distance to the straight lines drawn between Pmin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{min}}$$\end{document} and Plin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{lin}}$$\end{document} or Plin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{lin}}$$\end{document} and Pmax\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{\textrm{max}}$$\end{document}, respectively
… 
The normalised empirical kernel h~\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{h}$$\end{document} for all data of each of the three riders, separately
The normalised empirical kernel h~\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{h}$$\end{document} for all data of each of the three riders, separately
… 
The normalised kernel h~\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{h}$$\end{document} for all data of Rider A for our two approaches. The solid line corresponds to the empirical kernel after performing multiple linear regression for modeling the heart rate at time t with the pedal power at time t and previous times as predictor variables. The exponential decaying function is displayed by the dashed line
The normalised kernel h~\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{h}$$\end{document} for all data of Rider A for our two approaches. The solid line corresponds to the empirical kernel after performing multiple linear regression for modeling the heart rate at time t with the pedal power at time t and previous times as predictor variables. The exponential decaying function is displayed by the dashed line
… 
Median of the 99th percentiles for the distributions of the convolved pedal power (CPP) for all days in our data collection as a function of the accumulation window. For each rider, we normalised the values with respect the value of the median for an accumulation window of 14 days. Here, we consider a recuperation time τ=40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau = 40$$\end{document} seconds and a convolution window of 120 s
+6
Median of the 99th percentiles for the distributions of the convolved pedal power (CPP) for all days in our data collection as a function of the accumulation window. For each rider, we normalised the values with respect the value of the median for an accumulation window of 14 days. Here, we consider a recuperation time τ=40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau = 40$$\end{document} seconds and a convolution window of 120 s
… 
Figures - available from: Data Mining and Knowledge Discovery
This content is subject to copyright. Terms and conditions apply.
A preview of this full-text is provided by Springer Nature.
Learn more
Content available from Data Mining and Knowledge Discovery
This content is subject to copyright. Terms and conditions apply.
Data Mining and Knowledge Discovery (2023) 37:1125–1153
https://doi.org/10.1007/s10618-022-00905-5
Exploiting sensor data in professional road cycling:
personalized data-driven approach for frequent fitness
monitoring
Arie-Willem de Leeuw1·Mathieu Heijboer2·Tim Verdonck1·
Arno Knobbe3·Steven Latré1
Received: 22 April 2022 / Accepted: 29 November 2022 / Published online: 16 December 2022
© The Author(s), under exclusive licence to Springer Science+Business Media LLC, part of Springer Nature 2022
Abstract
We present a personalized approach for frequent fitness monitoring in road cycling
solely relying on sensor data collected during bike rides and without the need for
maximal effort tests. We use competition and training data of three world-class cyclists
of Team Jumbo–Visma to construct personalised heart rate models that relate the heart
rate during exercise to the pedal power signal. Our model captures the non-trivial
dependency between exertion and corresponding response of the heart rate, which we
show can be effectively estimated by an exponential kernel. To construct the daily
heart rate models that are required for day-to-day fitness estimation, we aggregate
all sessions in the previous week and apply sampling. On average, the explained
variance of our models is 0.86, which we demonstrate is more than twice as large as for
models that ignore the temporal integration involved in the heart’s response to exercise.
We show that the fitness of a cyclist can be monitored by tracking developments of
parameters of our heart rate models. In particular, we monitor the decay constant
of the kernel involved, and also analytically determine virtual aerobic and anaerobic
thresholds. We demonstrate that our findings for the virtual anaerobic threshold on
average agree with the results of exercise tests. We believe this work is an important
step forward in performance optimization by opening up avenues for switching to
adaptive training programs that take into account the current physiological state of an
athlete.
Responsible editor: Ulf Brefeld and Albrecht Zimmermann
Mathieu Heijboer, Tim Verdonck, Arno Knobbe and Steven Latré have contributed equally to this work.
BArie-Willem de Leeuw
arie-willem.deleeuw@uantwerpen.be
1University of Antwerp - imec, Sint-Pietersvliet 7, 2000 Antwerp, Belgium
2Team Jumbo-Visma, Het Zuiderkruis 23, 5215, MV ’s-Hertogenbosch, The Netherlands
3LIACS, Leiden University, Niels Bohrweg 1, 2333, CA Leiden, The Netherlands
123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
... The second key consideration is that, while HR is one of the most used physiological responses in the training context (Buchheit, 2014), and numerous studies have demonstrated the use of modelling HR dynamics for athlete monitoring purposes (Fujihara et al., 1973;Gielen et al., 2022;Lefever et al., 2014;Mongin et al., 2020aMongin et al., , 2020bStirling et al., 2008;Wang & Hunt, 2021;Zakynthinaki, 2015), no kayaking-specific research has been presented on the topic of modelling HR responses whereas numerous studies exist for other sports. For instance, HR dynamics have been linked to performance indicators in cycling (de Leeuw et al., 2023;Mongin et al., 2020aMongin et al., , 2020b. Furthermore, HR modelling studies have been performed for running (De Smet et al., 2016) and volleyball (Gielen et al., 2022), as well as for other physiological responses such as VO 2 (Poole & 2 -JASPER ET AL. ...
Modelling heart rate dynamics in relation to speed and power output in sprint kayaking as a basis for training evaluation and optimisation
Article
Full-text available
With the development of power output sensors in the field of paddle sports and the ongoing advancements in dynamical analysis of exercise data, this study aims to model the measurements of external training intensity in relation to heart rate (HR) time‐series during flat‐water kayak sprint. Nine elite athletes performed a total of 47 interval training sessions with incremental intensity (light to (sub‐) maximal effort levels). The data of HR, speed and power output were measured continuously and rating of perceived exertion and blood lactate concentration ([BLa]) were sampled at the end of each interval stage. Different autoregressive‐exogenous (ARX) modelling configurations are tested, and we report on which combination of input (speed or power), model order (1st or 2nd), parameter estimation method (time‐(in)variant) and training conditions (ergometer or on‐water) is best suited for linking external to internal measures. Average model R² values varied between 0.60 and 0.97, with corresponding average root mean square error values of 15.6 and 3.2 bpm. 1st order models with time‐varying (TV) parameter estimates yield the best model performance (average R² = 0.94). At the level of the individual athlete, the TV modelling features (i.e., the model parameters and derivatives such as time constant values) show significant repeated measure correlations in relation to measures of exercise intensity. In conclusion, the study provides a comprehensive description of how the dynamic relationship between external load and HR for sprint kayaking training data can be modelled. Such models can be used as a basis for improving training evaluation and optimisation.
View
... However, such an anonymized data set cannot be created easily as the information on the ending position could easily lead to linkage attacks [5], which would lead to the unauthorized sharing of personal information on these athletes. Moreover, researchers cannot use truly personal data without explicit permission of athletes, leading to limited data samples of a few athletes [8] or researchers using fully publicly available sources. A considerable downside, however, is that these data sources are scattered around the internet, leading to studies which use different data sources [9][10][11]. ...
Geospatial Road Cycling Race Results Data Set
Preprint
Full-text available
  • Sep 2024
The field of cycling analytics has only recently started to develop due to limited access to open data sources. Accordingly, research and data sources are very divergent, with large differences in information used across studies. To improve this, and facilitate further research in the field, we propose the publication of a data set which links thousands of professional race results from the period 2017-2023 to detailed geographic information about the courses, an essential aspect in road cycling analytics. Initial use cases are proposed, showcasing the usefulness in linking these two data sources.
View
... However, the tests for measuring it are arduous and performing them can disrupt an athlete's training such that it cannot be measured often. Thus, there is work on trying to predict this value based on less strenuous activities (de Leeuw et al., 2023;De Brabandere et al., 2018). Finally, teams may want to estimate parameters (e.g., total distance, sprint distance, number of accelerations) from players Content courtesy of Springer Nature, terms of use apply. ...
Methodology and evaluation in sports analytics: challenges, approaches, and lessons learned
Article
Full-text available
  • Jul 2024
  • MACH LEARN
There has been an explosion of data collected about sports. Because such data is extremely rich and complex, machine learning is increasingly being used to extract actionable insights from it. Typically, machine learning is used to build models and indicators that capture the skills, capabilities, and tendencies of athletes and teams. Such indicators and models are in turn used to inform decision-making at professional clubs. Designing these indicators requires paying careful attention to a number of subtle issues from a methodological and evaluation perspective. In this paper, we highlight these challenges in sports and discuss a variety of approaches for handling them. Methodologically, we highlight that dependencies affect how to perform data partitioning for evaluation as well as the need to consider contextual factors. From an evaluation perspective, we draw a distinction between evaluating the developed indicators themselves versus the underlying models that power them. We argue that both aspects must be considered, but that they require different approaches. We hope that this article helps bridge the gap between traditional sports expertise and modern data analytics by providing a structured framework with practical examples.
View
... The real-time capabilities of IoT technology enable real-time data processing [40], [41], [42]. The mechanism for real-time data processing and feedback is depicted in Figure 4. ...
The Analysis of Communication Strategy of Disabled Sports Information Based on Deep Learning and the Internet of Things
Article
Full-text available
  • Jan 2024
The ever-growing landscape of Internet of Things (IoT) technology and the evolution of deep learning algorithms have ushered in transformative changes in the communication strategy for disseminating information on disabled sports. This specialized information resource aims to provide relevant support and services related to sports activities for disabled individuals. This study investigates the communication strategy of disabled sports information driven by deep learning within the framework of the IoT and assesses the practical application performance of the proposed model. To achieve this objective, an appropriate deep learning model for the dissemination of sports information for the disabled is selected through a thorough literature review. Subsequently, an experimental framework is proposed for comprehensive performance verification, evaluating the model’s performance in reasoning time and user satisfaction through comparative experiments. By constructing deep learning models, extensive data on disabled sports activities are analyzed, enabling the identification and prediction of key factors in information dissemination. The results indicate that the proposed sports information dissemination model outperforms similar models across various performance metrics, particularly in real-time performance and user experience. Comparative analysis with attention-based deep neural networks and traditional machine learning algorithms reveals that the proposed model achieves an accuracy rate as high as 0.85, significantly surpassing the 0.78 and 0.82 accuracies of these models, respectively. Moreover, the proposed model demonstrates the shortest inference time (15ms), surpassing both aforementioned models. This study validates the relative advantages of the proposed model through comparison with similar studies, offering a novel solution for the dissemination of sports information for the disabled.
View
... knowledge discovery. To collect this helpful information, it is crucial to design strategies and construct data storages [1]. Data mining (DM) is a technology for finding knowledge in unstructured data collections that was not previously recognized. ...
Analysis of Social Network Data Mining for Security Intelligence Privacy Machine Learning
Article
  • Aug 2023
The Modern communication on the Internet platform is most responsive through social media. Social media has changed and is still reshaping how we share our thoughts and emotions in communication. It has introduced a constant real-time communication pattern that was before unheard of. Young and old, organizations, governmental agencies, professional associations, etc., all have social media accounts that they use exclusively for communication with other users. Social media also acts as a powerful network engine that connects users regardless of where they are in the world. The development of global communication will greatly benefit from the availability of this new communication platform in the future. Consequently, there is a pressing need to research usage trends. Therefore, it is vital to investigate social media platform usage trends in order to develop automated systems that intelligence services can use to help avert national security incidents. Through the use of social media data mining, this research study suggests an automated machine learning model that can improve speedy response to crises involving national and International security.
View
PerfoRank: cluster-based performance ranking for improved performance evaluation and estimation in professional cycling
Article
Full-text available
  • Jan 2025
  • MACH LEARN
Current cycling analytics solutions do not account for the race course profile or the level of the competition. Therefore, this paper develops a unique two-stage clustering-based ranking approach for rider evaluation. Initially, races are segmented into coherent clusters based upon elevation and road surface type. Subsequently, underlying skill levels are determined per cluster through the observed race results using the TrueSkill algorithm which allows to model multi-entrant competitions. The results indicate that our approach uncovers clusters which match the commonly known specializations in road cycling. The ranking methodology generates skill ratings which enable the identification of specialization and can be used in downstream tasks. Our results show that the proposed rankings drastically improve race outcome estimation when adding these rankings as features to the current prediction models.
View
Performance Measurement 2.0: Towards a Data-Driven Cyclist Specialization Evaluation
Conference Paper
  • Jan 2024
Current cycling analytics solutions do not account for either the raced course profile or the level of the competition. Therefore, this paper suggests a two-stage approach which initially clusters races into coherent clusters based upon both elevation and road surface type. Subsequently, underlying skill levels are determined per cluster through the observed race results. Our results indicate that the methodology results into clusters which match the commonly known specializations in road cycling. The ranking methodology results into skill ratings which enable the identification of specialization and can be used in other downstream tasks.
View
Boat Speed Prediction in SailGP
Conference Paper
  • Feb 2024
The significance of data analysis in high-performance sports has largely increased in recent years offering opportunities for further exploration using machine learning techniques. As a pioneer work in the academic community, our work showcases the power of data-driven approaches in enhancing performance and decision-making at high-performance sailing events. Specifically, we explore the application of data mining techniques on the dataset collected at a high-performance sailing event in Bermuda in 2021. By analyzing data from Race 4, the study aims to gain valuable insights into the relationship between variables such as wind speed, wind direction, foil usage, and daggerboard adjustments, and their impact on boat speed. Various prediction models, including Gradient Boosting, Random Forest, and a stacked model, were employed and evaluated using performance metrics like R2 score and mean squared error. The results demonstrate the models’ ability to accurately predict boat speed. These findings can be utilized to refine race strategies, optimize sail and rudder settings, and improve overall performance in SailGP races. Future plans include collaboration with SailGP to work with larger datasets and integrate the models into live racing scenarios.
View
The relationships between age and race performance in women’s road cycling
Article
  • Jan 2024
We investigate the relationships between age and race performance in women’s road cycling. Therefore, we consider public data from ProCyclingStats (PCS) on race results of 558 female riders present in the top 500 of one of the yearly rankings between 1993 and 2021. We apply a data-driven approach consisting of three subsequent steps. First, we train decision tree classifiers to distinguish between race profiles. Second, we cluster the riders based on the fraction of PCS points obtained in different race types and uniquely identified the speciality of a rider as Sprinter, All Rounder or One Day. Third, we use polynomial regression to determine relationships between age and a rider’s yearly race performance, defined as the average number of points per race entered in a single year, for each cluster. We obtain that the yearly performance is independent of age for one-day specialists. For sprinters and all-rounders, we found that the race performance increases until it reaches a maximum at 28.7 (sprinters) and 27.0 (all-rounders) years and then decreases again. Our findings can be used as a benchmarking tool for monitoring the development of women’s road cyclists, talent identification or constructing long-term training programmes.
View
An Application On Personnel Selection in Human Resources
Conference Paper
Full-text available
  • Sep 2023
zet: Etkili işe almanın kurumsal başarıdaki kilit rolünü kabul eden araştırma, hem geleneksel hem de en ileri teknolojiye sahip çok çeşitli seçim metodolojilerini titizlikle inceliyor (Edwards ve diğerleri, 2022). Kapsamlı bir veri kümesinden yararlanan bu çalışma, mevcut uygulamalara, zorluklara ve personel seçiminin gelecekteki yönüne ışık tutmak için nitel vaka çalışmaları ve nicel veri analitiğinden yararlanan karma yöntemli bir yaklaşım kullanır (Avrahami ve diğerleri, 2022). Araştırma, işe alım uygulamalarında değişikliklere yol açan sosyo-teknolojik faktörleri daha da derinlemesine araştırıyor. Dijital platformların ortaya çıkışı, küreselleşen yetenek havuzları ve uzaktan çalışma uyumluluğuna duyulan acil ihtiyaç, İK profesyonellerini yenilik yapmaya ve uyum sağlamaya zorladı. Bu çalışma aynı zamanda adaylarda teknik yeterliliklerinin yanı sıra kültürel uyum, sosyal beceriler ve uyum sağlama yeteneğinin artan önemini vurgulamaktadır (Pessach ve diğerleri, 2020). Kuruluşlar coğrafi sınırları aşıp dijital alemin derinliklerine daldıkça, 'ideal aday' tanımı sürekli olarak yeniden tanımlanıyor ve bu da gelişmiş ve çeşitlendirilmiş seçim metodolojilerini garanti ediyor. Ayrıca araştırma, işe alım sürecinde geri bildirim döngülerinin önemini vurgulamaktadır. İşe alım stratejilerinde sürekli öğrenme ve yinelemeli iyileştirme fikri çok önemlidir. Kuruluşlar, işe alım sonrası verileri analiz ederek seçim metodolojilerinin etkililiğini belirleyebilir ve gelecekteki işe alımlar için gerekli ayarlamaları yapabilir. Bu, yalnızca işe alım sürecinin ilgili kalmasını sağlamakla kalmaz, aynı zamanda çalışan devrini azaltmaya yardımcı olur ve böylece kuruluş için önemli maliyet ve kaynak tasarrufu sağlar (Chowdhury ve diğerleri, 2023). Çalışma, İK profesyonellerine, yalnızca yetkin değil, aynı zamanda dirençli ve kurumsal değerlerle uyumlu bir işgücü oluşturmak için işe alımın hem teknolojik ilerlemelerini hem de insan merkezli yönlerini benimseyen sürekli öğrenmeye katılmaları için bir eylem çağrısı ile sona eriyor. Abstract: Recognizing the pivotal role of efficient hiring in organizational success, the research rigorously examines a wide array of selection methodologies, both traditional and cutting-edge (Edwards et al., 2022). Drawing from an extensive dataset, this study employs a mixed-method approach, utilizing qualitative case studies and quantitative data analytics to shed light on current practices, challenges, and the future direction of personnel selection (Avrahami et al., 2022). The research further probes into the socio-technological factors that have driven changes in hiring practices. The emergence of digital platforms, globalized talent pools, and the pressing need for remote work adaptability have compelled HR professionals to innovate and adapt. This study also highlights the growing importance of cultural fit, soft skills, and adaptability in candidates, alongside their technical competencies (Pessach et al., 2020). As organizations transcend geographical boundaries and dive deeper into the digital realm, the definition of an 'ideal candidate' is continually being redefined, warranting advanced and diversified selection methodologies. Moreover, the research emphasizes the significance of feedback loops in the hiring process. The idea of continuous learning and iterative improvement in recruitment strategies is paramount. By analyzing post-hiring data, organizations can ascertain the effectiveness of their selection methodologies, making necessary adjustments for future hires. This not only ensures that the hiring process remains relevant but also aids in reducing employee turnover, thereby saving considerable costs and resources for the organization (Chowdhury et al., 2023). The study concludes with a call to action for HR professionals to engage in perpetual 17 th UBAK, 19-20 August 2023, Ankara | 299 learning, embracing both the technological advancements and human-centric aspects of recruitment, to build a workforce that is not just competent but also resilient and aligned with the organizational ethos.
View
Deep reinforcement learning for improving competitive cycling performance
Article
Full-text available
  • May 2022
  • EXPERT SYST APPL
Developing expert systems that make use of artificial intelligence (AI) to provide predictive analytics as well as targeted recommendations for decision support has been gaining momentum in recent years. Both academia and industry are looking into creating such systems to solve real-world problems and tackle specific challenges. In our work, we investigate the potential application of different machine learning approaches to solutions around competitive cycling. Specifically, we build and evaluate prediction models that are capable of accurately predicting a cyclist’s speed and heart rate using sensory information collected during bike rides. In addition, we create a recommendation module that is able to provide real-time action suggestions to cyclists regarding their posture with the goal of improving their overall performance. We achieve this using a combination of model-based reinforcement learning (RL) and deep RL. In particular, we use model-based RL to learn a “simulator” of bike rides using the prediction models and action profiles extracted from sensors placed on the cyclists’ back. We then use deep Q-learning in the simulator to extract policies that improve a cyclist’s behavior during a bike ride. Our evaluation shows that by recommending specific actions throughout the ride, cyclists can increase their overall average speed with only a minimal impact on their heart rate. The results presented in this paper constitute clear evidence that advanced AI techniques are a prime candidate for further developing intelligent solutions in competitive cycling and other similar areas.
View
Predicting the next Pogačar: a data analytical approach to detect young professional cycling talents
Article
Full-text available
  • Jan 2022
  • ANN OPER RES
The importance of young athletes in the field of professional cycling has sky-rocketed during the past years. Nevertheless, the early talent identification of these riders largely remains a subjective assessment. Therefore, an analytical system which automatically detects talented riders based on their freely available youth results should be installed. However, such a system cannot be copied directly from related fields, as large distinctions are observed between cycling and other sports. The aim of this paper is to develop such a data analytical system, which leverages the unique features of each race and thereby focusses on feature engineering, data quality, and visualization. To facilitate the deployment of prediction algorithms in situations without complete cases, we propose an adaptation to the k-nearest neighbours imputation algorithm which uses expert knowledge. Overall, our proposed method correlates strongly with eventual rider performance and can aid scouts in targeting young talents. On top of that, we introduce several model interpretation tools to give insight into which current starting professional riders are expected to perform well and why.
View
Power profiling and the power-duration relationship in cycling: a narrative review
Article
Full-text available
  • Feb 2022
  • EUR J APPL PHYSIOL
Emerging trends in technological innovations, data analysis and practical applications have facilitated the measurement of cycling power output in the field, leading to improvements in training prescription, performance testing and race analysis. This review aimed to critically reflect on power profiling strategies in association with the power-duration relationship in cycling, to provide an updated view for applied researchers and practitioners. The authors elaborate on measuring power output followed by an outline of the methodological approaches to power profiling. Moreover, the deriving a power-duration relationship section presents existing concepts of power-duration models alongside exercise intensity domains. Combining laboratory and field testing discusses how traditional laboratory and field testing can be combined to inform and individualize the power profiling approach. Deriving the parameters of power-duration modelling suggests how these measures can be obtained from laboratory and field testing, including criteria for ensuring a high ecological validity (e.g. rider specialization, race demands). It is recommended that field testing should always be conducted in accordance with pre-established guidelines from the existing literature (e.g. set number of prediction trials, inter-trial recovery, road gradient and data analysis). It is also recommended to avoid single effort prediction trials, such as functional threshold power. Power-duration parameter estimates can be derived from the 2 parameter linear or non-linear critical power model: P ( t ) = W ′/ t + CP ( W ′—work capacity above CP; t —time). Structured field testing should be included to obtain an accurate fingerprint of a cyclist’s power profile.
View
A Learn-to-Rank Approach for Predicting Road Cycling Race Outcomes
Article
Full-text available
  • Oct 2021
Professional road cycling is a very competitive sport, and many factors influence the outcome of the race. These factors can be internal (e.g., psychological preparedness, physiological profile of the rider, and the preparedness or fitness of the rider) or external (e.g., the weather or strategy of the team) to the rider, or even completely unpredictable (e.g., crashes or mechanical failure). This variety makes perfectly predicting the outcome of a certain race an impossible task and the sport even more interesting. Nonetheless, before each race, journalists, ex-pro cyclists, websites and cycling fans try to predict the possible top 3, 5, or 10 riders. In this article, we use easily accessible data on road cycling from the past 20 years and the Machine Learning technique Learn-to-Rank (LtR) to predict the top 10 contenders for 1-day road cycling races. We accomplish this by mapping a relevancy weight to the finishing place in the first 10 positions. We assess the performance of this approach on 2018, 2019, and 2021 editions of six spring classic 1-day races. In the end, we compare the output of the framework with a mass fan prediction on the Normalized Discounted Cumulative Gain (NDCG) metric and the number of correct top 10 guesses. We found that our model, on average, has slightly higher performance on both metrics than the mass fan prediction. We also analyze which variables of our model have the most influence on the prediction of each race. This approach can give interesting insights to fans before a race but can also be helpful to sports coaches to predict how a rider might perform compared to other riders outside of the team.
View
Result-based talent identification in road cycling: discovering the next Eddy Merckx
Article
Full-text available
  • Oct 2021
  • ANN OPER RES
In various sports large amounts of data are nowadays collected and analyzed to help scouts with identifying talented young athletes. In contrast, the literature on result-based talent identification in road cycling is remarkably scarce. The purpose of this paper is to provide insight into the possibilities of the use of publicly available data to discover new talented Under-23 (U23) riders via statistical learning methods (linear regression and random forest techniques). At the same time, we try to find out the main determinants of success for U23 riders in their first years of professional cycling. We collect results for more than 25000 road cycling races from 2007–2018 and consider more than 2500 riders from over 80 countries. We use the data from 2007 to 2017 to train and validate our models, and use the data from 2018 to predict how well U23 riders will perform in their first three elite years. Our results reveal that past U23 race results appear to be important predictors of future cycling performance.
View
Personalized Machine Learning Approach to Injury Monitoring in Elite Volleyball Players
Article
Full-text available
We implemented a machine learning approach to investigate individual indicators of training load and wellness that may predict the emergence or development of overuse injuries in professional volleyball. In this retrospective study, we collected data of 14 elite volleyball players (mean ± SD age: 27 ± 3 years, weight: 90.5 ± 6.3 kg, height: 1.97 ± 0.07 m) during 24 weeks of the 2018 international season. Physical load was tracked by manually logging the performed physical activities and by capturing the jump load using wearable devices. On a daily basis, the athletes answered questions about their wellness, and overuse complaints were monitored via the Oslo Sports Trauma Research Center (OSTRC) questionnaire. Based on training load and wellness indicators, we identified subgroups of days with increased injury risk for each volleyball player using the machine learning technique Subgroup Discovery. For most players and facets of overuse injuries (such as reduced sports participation ), we have identified personalized training load and wellness variables that are significantly related to overuse issues. We demonstrate that the emergence and development of overuse injuries can be better understood using daily monitoring, taking into account interactions between training load and wellness indicators, and by applying a personalized approach. Highlights With detailed, athlete‐specific monitoring of overuse complaints and training load, practical insights in the development of overuse injuries can be obtained in a player‐specific fashion contributing to injury prevention in sports. A multi‐dimensional and personalized approach that includes interactions between training load variables significantly increases the understanding of overuse issues on a personal basis. Jump load is an important predictor for overuse injuries in volleyball.
View
Time Series Regression in Professional Road Cycling
Conference Paper
Full-text available
  • Oct 2020
With the recent explosive developments in sensoring capabilities and ubiquitous computing in road cycling, large quantities of detailed data about performance are becoming available. In this paper, we will demonstrate that this rich data in cycling offers several non-trivial data science challenges. The primary task that we focus on is a regression task: given a collection of results in previous races of a specific rider, predict the performance in a future race solely based on the characteristics of said rider and the stage profile. To make these predictions, we have developed a predictive pipeline that consists of three consecutive rider-specific models. First, we transform the distance-altitude profile into a time profile, by using a climb-descent model that describes the relationship between the speed of the cyclist and the slope of the terrain. Second, we introduce an effective profile that includes the rider-specific physiological capabilities. Third, we predict the performance based on the characteristics of the effective profile, by using a model constructed from the historical records of our cyclist. To demonstrate the relevance of this work, we show that for a professional cycling team, important information for making tactical decisions can be obtained from our modeling approach.
View
Sensor-Based Performance Monitoring in Track Cycling
Chapter
  • Apr 2022
Research has not found its way yet to the track cycling madison discipline. Currently, training files are collected from cycling computers, after which the data is interpreted in a mainly subjective manner, based on the domain knowledge of a coach. The goal of this paper is twofold. Starting with the automated detection of madison handslings from cadence, acceleration and gyroscope data, all other data corresponding to a single handsling can easily be obtained. The second goal concerns the calculation of statistics on rider performances during a handsling. We present two madison handsling performance assessment use cases. The first use case exposes imbalances within a madison rider pair, whereas the second use case employs power data to monitor the effort a single rider puts into the handsling.
View
Validity of Detrended Fluctuation Analysis of Heart Rate Variability to determine intensity thresholds in elite cyclists
Article
  • Mar 2022
Background: The evaluation of performance in endurance athletes and the subsequent individualization of training is based on the determination of individual physiological thresholds during incremental tests. Gas exchange or blood lactate analysis are usually implemented for this purpose, but these methodologies are expensive and invasive. The short-term scaling exponent alpha 1 of detrended Fluctuation Analysis (DFA-α1) of the Heart Rate Variability (HRV) has been proposed as a non-invasive methodology to detect intensity thresholds. Purpose: The aim of this study is to analyse the validity of DFA-α1 HRV analysis to determine the individual training thresholds in elite cyclists and to compare them against the lactate thresholds. Methodology: 38 male elite cyclists performed a graded exercise test to determine their individual thresholds. HRV and blood lactate were monitored during the test. The first (LT1 and DFA-α1-0.75, for lactate and HRV, respectively) and second (LT2 and DFA-α1-0.5, for lactate and HRV, respectively) training intensity thresholds were calculated. Then, these points were matched to their respective power output (PO) and heart rate (HR). Results: There were no significant differences (p > 0.05) between the DFA-α1-0.75 and LT1 with significant positive correlations in PO (r = 0.85) and HR (r = 0.66). The DFA-α1-0.5 was different against LT2 in PO (p = 0.04) and HR (p = 0.02), but it showed significant positive correlation in PO (r = 0.93) and HR (r = 0.71). Conclusions: The DFA1-a-0.75 can be used to estimate LT1 non-invasively in elite cyclists. Further research should explore the validity of DFA-α1-0.5.
View
Time Series Regression in Professional Road Cycling
Chapter
  • Oct 2020
  • Lect Notes Comput Sci
With the recent explosive developments in sensoring capabilities and ubiquitous computing in road cycling, large quantities of detailed data about performance are becoming available. In this paper, we will demonstrate that this rich data in cycling offers several non-trivial data science challenges. The primary task that we focus on is a regression task: given a collection of results in previous races of a specific rider, predict the performance in a future race solely based on the characteristics of said rider and the stage profile. To make these predictions, we have developed a predictive pipeline that consists of three consecutive rider-specific models. First, we transform the distance-altitude profile into a time profile, by using a climb-descent model that describes the relationship between the speed of the cyclist and the slope of the terrain. Second, we introduce an effective profile that includes the rider-specific physiological capabilities. Third, we predict the performance based on the characteristics of the effective profile, by using a model constructed from the historical records of our cyclist. To demonstrate the relevance of this work, we show that for a professional cycling team, important information for making tactical decisions can be obtained from our modeling approach.
View