Abstract: The multi-attribute intelligent risk assessment method of the public services for national fitness, based on the "improved entropy weight method-TOPSIS method-grey relational analysis" framework, adopted the 5M1E theory framework, combined with the random forest algorithm to screen risk indicators, to calculate the weight of evaluation indicators with the improved entropy weight method, including 24 indicators such as scientific fitness training (weight of 0.115), privacy protection degree (weight of 0.144), and 5G infrastructure coverage rate (weight of 0.072). TOPSIS method was used to rank risk events and grey correlation analysis to verify and correct the results. Based on the intelligent risk event of public services for national fitness in Taiyuan City, the research divides the intelligent risk indicators into four levels: mild (range: 0, 0.34)), moderate (range: 0.34, 0.39)), severe (range: 0.39, 0.45)), and special (range: 0.45, 1), and the each level's the characteristics and impacts are described in details. This empirical research may accurately reveals the potential risks in intelligent services, providing a scientific basis for developing targeted risk management strategies, and opening up a new path for risk assessment research in the field of intelligent public services.