Abstract：In order to solve the problems of single evaluation and low prediction accuracy in the current ecological environment service value system, this paper puts forward three dynamic combination
forecasting systems for ecosystem service value, namely ARI-
MA-BP, PCABP-ARIMA and PCABP ARIMA prediction model, and makes an empirical analysis on the ecosystem service value
prediction of grassland in Tibet Autonomous region, forest land in Heilongjiang Province and cultivated land in Henan Province.
The experimental results show that the average absolute error of based on ARIMA-BP ecosystem service value prediction model is
only 0.87%, and the average absolute error of PCABP-ARIMA and PCABP+ARIMA prediction model is 1.98% and 2.24% respec-
tively.ARIMA-BP is the best, which is 62.67% smaller than the traditional BP neural network prediction model
