Traditional quantitative analysis methods have limitations in dealing with large-scale data and complex economic relations. Therefore, this paper proposes an optimization framework based on the combination of dynamic state space model and intelligent optimization algorithm. In the data preprocessing stage, we use stationarity test, cointegration analysis and principal component analysis (PCA) to reduce dimensions to process economic time series data. In the core modeling stage, a dynamic state space model is constructed and a time-varying parameter mechanism is introduced. In the optimization stage, the improved genetic algorithm (GA) is used to optimize the parameters globally, and a hybrid Kalman filter-particle filter (Kalman-PF) joint estimation algorithm is designed. Through empirical research, taking the macroeconomic data of China from 2000 to 2020 as a sample, the results show that the root mean square error (RMSE) of the optimized model is reduced from 0.87 to 0.68, the average absolute error (MAE) is reduced from 0.62 to 0.49, and the prediction time is shortened from 4.2 s to 2.6 s, and the parameter stability of the optimized model is obviously better than that of the benchmark model, showing stronger prediction accuracy.