Authors :

Koo Li Shan and Nurul Izzati Rahman

Address :

Department of Software Systems, Universiti Putra Malaysia, Malaysia

Department of Data Analytics, International Islamic University Malaysia, Malaysia

Abstract :

Research in sports analytics has mostly concentrated on measuring and analyzing performance in three contexts: training, competition, and recovery. Recognizing and analyzing athlete behavior is crucial for enhancing performance, making the most of training, and preventing injuries in sports. The complex and irregular patterns in athletic performance are beyond the capabilities of conventional approaches and mainly depend on subjective observation and oversimplified statistical analysis. This work presents SA-EAMDL, a cutting-edge Sports Analytics (SA) system that combines an Evolutionary Algorithm (EA) with a Meta-Deep Learning (MDL) model to identify the actions of athletes accurately. Its purpose is to address these difficulties. To achieve consistently high accuracy across different datasets, the evolutionary method optimizes the structure and hyperparameters of a deep learning model. Using convolutional neural networks (CNNs) and recurrent neural networks (RNNs), the meta-deep learning model analyzes data on the actions of athletes and can concurrently recognize complex patterns over time and place. The accuracy and efficiency of behavior recognition in sports should be greatly improved by integrating these technologies. This approach is well-suited to modern sports analytics as it reduces the computing load of testing and training models and improves the accuracy of researching player behaviour. Athletes in any activity can benefit from this method's capacity to assist them make quicker, more data-driven decisions on performance evaluations, practice efficacy, and prevention of injuries strategies. The SA-EAMDL framework can greatly improve sports analytics due to its effectiveness and precision.

Keywords :

Meta Deep Learning model, Evolutionary Algorithm, Convolutional Neural Network, Recurrent Neural Networks, Sports Analytics, Athletic Behaviour.