The total dependency ratio of Shanghai is affected by many factors, leading to volatile data, it can not be predicted more accurately by gray prediction model, so the article proposed an improved GM(1, 1) model. Will improve the precision of the model were compared with the original model obtained using quadratic polynomial fitting improved model prediction precision higher than the original model, so quadratic polynomial fitting GM(1, 1) improved model is adopted. And use this improved model into the prediction of the total dependency ratio of Shanghai, and combined total dependency ratio data from 2007 to 2011 in Shanghai, a new forecasting model is built, and the model with the 2012 data to verify eligibility can be used to predict the total dependency ratio of Shanghai next few years to facilitate the economic development of the city of the future macro. The results show that the prediction method is reasonably practicable, and provides a theoretical basis for other related forecasts.