This paper applies a data-driven dynamic transmission rate to replace the basic reproduction number $R_0$ and studies the characteristics and trends of the development of COVID-19 at both national and provincial levels. Firstly, based on the dynamic growth rate, an ordinary differential equation for infectious diseases is established, which can derive the dynamic transmission rate model. Secondly, this paper selects the power function as the fitting function of the dynamic transmission rate, and uses 3 days as the optimal sliding window period to estimate the inflection points in different regions. Finally, using the dynamic model, this paper predicts the starting point of the end phase of the epidemic at different levels in various places, and then compares and analyzes 9 epidemic-related indicators among 13 provinces and cities. The results show that the dynamic transmission rates in all regions have steadily declined after a brief fluctuation, which means the epidemic situation has been effectively controlled; the date of the estimated inflection points are mainly concentrated in mid-February, and the predicted end phase will come before the end of March; at the same time, there are some differences in the characteristics and trends of the epidemic situation as well as the intensity and effectiveness of prevention and control measures.