The current research on the long-term settlement prediction of tunnels under subway loading is generally based on the maximum passenger loading condition, and the influence of passenger flow changes on tunnel settlement is seldom taken into account, which leads to the discrepancy between the calculation conditions and the actual working conditions. In this regard, firstly, from the perspective of passenger flow pattern, the metro operation is divided into three working conditions, namely "peak period, secondary period and low period" by using data mining method. Afterwards, the dynamic response characteristics of the soil around the tunnel under each working condition are compared by numerical simulation, and it is found that the dynamic response results of each passenger working condition are quite different, and the influence of the passenger flow changes on the tunnel settlement cannot be ignored in the study. Then on this basis, based on the principle of calculus, considering the change rule of metro passenger flow, the explicit integral prediction model of long-term tunnel settlement is established, and relying on the measured data of Shanghai Metro Line 1, the reliability of the model is verified. Finally, the long-term settlement characteristics of tunnels in silty soil areas are analysed by taking Shanghai Metro Line 10 as an example. The results indicate that under the same conditions, the soil displacement and dynamic bias stress around the tunnel increase with the passenger capacity. The improved integral prediction model yields more accurate results compared to the conventional model and can be applied to research related to the prediction of long-term settlement of the metro. The long-term settlement of Shanghai Metro Line 10 develops faster in the early stage of operation, and the settlement after 5 years of operation is about 30.53 mm, which is 82% of the total settlement after 20 years (37.07 mm).