ITS4U : Intelligent Tracking System for You

Abstract

The employment of digital technologies within the sporting domain has gained popularity in these last decades in an attempt to take advantage of the efficient processing and effective results produced, as well as exploit the additional information that smart processing of generated data can potentially provide. Tracking athletes during a marathon and predicting their finishing time is an arduous task. The data harnessed from such events can be adapted to better understand the athletes’ capabilities, preferences, and ambitions. In this project, we investigate different methods of how to optimise the use of the data retrieved from marathon events through the use of specialised RFID hardware and software. The use of an RFID setup, including an RFID reader, antenna, and tags, enables the detection and tracking of athletes during a marathon race. A marathon setup includes an input zone, a number of transition zones, and a destination zone. Each athlete is provided a wristband with an RFID tag containing the corresponding details. Every time an athlete passes through a checkpoint, their details are stored in a database. We make use of three different regression techniques to optimise the data acquired from a marathon race. The techniques include the use of Multiple Linear Regression, Gradient Boosting Regression, and Elastic Net Regression. A Naive Classifier is used so as to set a benchmark for comparison with the other regression models. In order to evaluate these techniques, three different metrics are utilised, including the Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R 2 ). When comparing these metrics, the Multiple Linear model was found to obtain the least error metric. However, all models manage to achieve a very low error metric. The Wall problem is also catered for through a smart processing of the slowdown effect in marathon runners during a race and comparing the interval distances together.