نبذة مختصرة : Trail running is a running sport that spans across outdoor and mountainous terrain, often with hilly ascents and descents. This created a competitive scene that has garnered the interest of trail running enthusiasts into competing in 100km+ grueling races that can last for over 24 hours. Machine learning is a research field that focuses on developing algorithms capable of learning and making predictions based on data. In essence, it revolves around creating algorithms that can analyze data, identify patterns, and make informed predictions. The organizations that setup trail running events operate a system that keeps track of the competitors arrival times across the designated points in a trail running course. The data analysis of trail running can provide valuable insights in respect to the planning of the event, namely in regard to the guarantee of the runner’s safety, the scheduling of transport for runners, the allocation of human resources, and the allocation of food and beverages. A first initial research finds that the existing literature does not focus on the prediction in regards to both trail running and machine learning topics, the latter being a solution found in research related to marathon running. At first, an initial approach was conducted in order to comprehend the prediction of arrival times based on an existing implementation. This approach analyzed the velocities of the runners, and such approach revealed that it would be optimal to utilize the Mean Absolute Error (MAE) as a metric for the next approaches. The second approach evolved into the use of the models LASSO and Random Forests with features related to both the checkpoints of a race and it’s runners in competition. This approach netted a 24.45% MAE reduction when compared to the first approach. In the third approach, the LASSO model was excluded as the Random Forest model had overcome with least MAE. With the inclusion of more race time data in the last approach, we were able to reduce the MAE of the first approach by 28.71% with the ...
No Comments.