The accurate prediction of the trend of real estate prices is an important content of the national monetary policy and economic regulation. It needs to construct the fluctuation of interest rate, the real estate investment and real estate prices for real estate price related variables, and the real estate price index dynamic trend is predicted. An improved real estate price correlation prediction algorithm is proposed based on maximum conjugate gradient continuous functional, predictive factors of the real estate prices are analyzed, the association SVM model is constructed with real estate prices, interest rates and real estate investment. PCA system design is obtained for real estate price estimation, the conjugate gradient maximum continuous function is proposed, singular semi positive definite double periodic is taken, and the prediction algorithm is obtained. The contribution degree of the effect of interest rate is taken into consideration, the price related characteristics of state information fusion is processed, and the prediction error can be reduced. The simulation results show that the prediction accuracy is higher, error is controlled within 1.5%, it has superior performance.