The impact of the COVID-19 lockdown on air quality in seven major Chinese cities was investigated by utilizing long-term datasets of air pollutants and meteorological conditions from 2016 to 2021. Generalized additive model (GAM) was developed to predict air quality during the lockdown period. The model accounting for weather conditions demonstrated high accuracy with predictions compared against measurements during the lockdown. Significant reductions in NO₂, CO, and PM₁₀ concentrations were observed primarily due to decreased vehicular traffic and industrial activities. Notable reductions were particularly evident in cities with high traffic volumes and industrial emissions prior to the lockdown. The study also employed transfer learning to enhance the accuracy of lockdown model with limited data. Despite occasional anomalies caused by specific events like fireworks and agricultural burning, the findings suggest that extended training periods and advanced modeling techniques can significantly improve air quality predictions. This research highlights the potential long-term benefits of sustained reductions in human activities and provides valuable insights for future air quality management and policy-making.