With the development of the economy at hand, local fiscal revenue forecasting is becoming increasingly important for policy making and budgetary management. In this paper, based on SVR (Support Vector Regression) model, the method of building a more accurate local fiscal revenue forecasting model is discussed. The study is constructing a local fiscal revenue forecasting model based on the SVR model to improve forecasting accuracy. The PCA method is applied to preprocess the data to reduce multicollinearity and extract the key influence factors. The SVR model is used to forecast, and the appropriate kernel function and model parameters are chosen. The model’s ability to predict accurately is demonstrated by the average prediction error of 2.47% in the short-term and 3.29% in the medium- and long-term forecasts. The PCA-SVR model is more effective in dealing with seasonal and short-term shocks than traditional time series models. The study proved that the SVR model is an effective and powerful tool for predicting local fiscal revenues.