THE SPATIO-TEMPORAL EVOLUTION OF OLYMPIC MEDALS BASED ON MULTILEVEL REGRESSION ANALYSIS AND MULTIDIMENSIONAL FEATURE COUPLING

Abstract

Olympic medal performance is a measure of a country’s competitive sport achievements but also largely shaped by various socioeconomic aspects. This study attempts to comprehensively investigate the spatiotemporal development of Olympic performance and to build a highly accurate hierarchical forecasting model. Firstly, using data from the 2008-2024 Olympic Games (five editions) after data cleaning and feature engineering it can be observed an extremely positive correlation between size of the games and total number of medals with correlation coefficient. The exploratory analysis suggests that GDP is the main socio-economic factor determining a nation’s medal competitiveness and its relationship with the number of medals has significant structural differences. It can be seen that for high-income countries, the conversion efficiency of economic resources into competitive advantages is best. This research devised separate forecast systems for nations on varying levels of competition; it used a share-smoothing prediction method on top-performing countries so grasp long-term trend; added up weight least-squares shares regression plus reliability shrinkage system for mid-ranking countries; and innovatively developed a two-step prediction framework integrating logistic regression and random forests models for those nations without any previous medal record. Forecast results suggest at the 2028 Los Angeles Olympics USA will get 150 medals because they are the host nation while China UK and France which are all powerhouses will stay in the lead positions. With the coupling of multiple features, this paper gives scientific evidence for people to know about the change in Olympic performance.