This study proposes a new hybrid model combining long short-term memory network and support vector regression algorithm to test the accuracy of economic indicator forecasting. In the past, traditional economic indicator forecasting methods, autoregressive integrated moving average model and SVR method, usually cannot effectively handle nonlinear characteristics and complex time dependencies in relevant data. LSTM algorithm can capture long-term and short-term dependencies in data through gating mechanism, which is more suitable for time series data analysis, while SVR maps data to high-dimensional space through kernel function, which can better capture nonlinear relationships and optimize the performance of regression model. Therefore, combining the advantages of both, the LSTM-SVR hybrid model proposed in this study has significant advantages in the task of economic indicator forecasting, especially when dealing with real key economic indicators such as GDP growth rate, inflation rate, and unemployment rate. This study conducted multiple evaluations on the model, including analysis of indicators such as mean square error (MSE) and mean absolute percentage error. Experimental results show that compared with the traditional benchmark model, the prediction accuracy and reliability of the LSTM-SVR model are significantly improved. The potential application of the model in actual economic decision-making is further analyzed, emphasizing its importance in policy-making and economic planning.