Improving Machine Learning Results for Detection of Weightlifting Injuries in Master Athletes
Oral Presentation
Paper ID : 1124-SSRC
Authors
دانشگاه کردستان
Abstract
Background:
Master athletes often experience sport-related injuries, which can have a substantial impact on their performance and overall well-being. Understanding the types and causes of these injuries is crucial for developing effective preventive measures. This research aimed to develop a machine learning model to identify injury risk factors.
Purpose:
This research seeks to identify and analyze sport injuries in master athletes, with the goal of developing targeted algorithms to predict the injuries, and so reduce the risk of injuries and enhance athletic performance.
Methodology:
The dataset utilized in this study, "Weightlifting Injuries in Master Athletes" [1], consists of 976 samples with 38 features. The features characterize various injury types across 5 different targets. As a necessary preprocessing step, relevant transformations were applied on the raw dataset. Specifically, a graph-based conversion approach was implemented by thresholding inter-feature correlations at 0.73. This graph representation permitted the extraction of an additional 12 features, resulting in an enhanced dataset containing a total of 50 features for each sample. Through this methodology, the richness of information embedded in the initial 976 samples has been more fully captured. To ensure rigorous benchmarking of all data analytics conducted herein, 10-fold cross-validation was systematically integrated. Several standard machine learning models, including Logistic Regression, Support Vector Machines (SVMs), Random Forests, and others, were trained on this dataset. By leveraging cross-validation alongside classical and state-of-the-art machine learning algorithms, the reliability of results and conclusions presented throughout this work has been critically substantiated.
Results:
The Logistic Regression algorithm achieved significant accuracies of 80.5% for Back, 79.4% for Knees, and 80.3% for Wrist, surpassing the benchmarks of existing studies [1] in the field. The performance highlights the proposed methodology’s efficacy in identifying and managing sports injuries in master athletes.
Conclusions:
The findings of this study demonstrate the potential of advanced data analysis and corrective exercise in enhancing injury prevention strategies for master athletes. By leveraging these insights, tailored interventions can be developed to mitigate the risk of sport injuries and promote the long-term well-being of master athletes.
Master athletes often experience sport-related injuries, which can have a substantial impact on their performance and overall well-being. Understanding the types and causes of these injuries is crucial for developing effective preventive measures. This research aimed to develop a machine learning model to identify injury risk factors.
Purpose:
This research seeks to identify and analyze sport injuries in master athletes, with the goal of developing targeted algorithms to predict the injuries, and so reduce the risk of injuries and enhance athletic performance.
Methodology:
The dataset utilized in this study, "Weightlifting Injuries in Master Athletes" [1], consists of 976 samples with 38 features. The features characterize various injury types across 5 different targets. As a necessary preprocessing step, relevant transformations were applied on the raw dataset. Specifically, a graph-based conversion approach was implemented by thresholding inter-feature correlations at 0.73. This graph representation permitted the extraction of an additional 12 features, resulting in an enhanced dataset containing a total of 50 features for each sample. Through this methodology, the richness of information embedded in the initial 976 samples has been more fully captured. To ensure rigorous benchmarking of all data analytics conducted herein, 10-fold cross-validation was systematically integrated. Several standard machine learning models, including Logistic Regression, Support Vector Machines (SVMs), Random Forests, and others, were trained on this dataset. By leveraging cross-validation alongside classical and state-of-the-art machine learning algorithms, the reliability of results and conclusions presented throughout this work has been critically substantiated.
Results:
The Logistic Regression algorithm achieved significant accuracies of 80.5% for Back, 79.4% for Knees, and 80.3% for Wrist, surpassing the benchmarks of existing studies [1] in the field. The performance highlights the proposed methodology’s efficacy in identifying and managing sports injuries in master athletes.
Conclusions:
The findings of this study demonstrate the potential of advanced data analysis and corrective exercise in enhancing injury prevention strategies for master athletes. By leveraging these insights, tailored interventions can be developed to mitigate the risk of sport injuries and promote the long-term well-being of master athletes.
Keywords