To address the key technical challenges in turboshaft engine performance modeling， such as complex operating conditions， difficulty in effectively distinguishing steady-state and transient data， and interference of maintenance activities on degradation modeling， a gas path performance prediction method for turboshaft engines based on test run data is proposed. First， Pearson correlation analysis is adopted to reduce the dimensionality of test run data， eliminating redundant information to improve modeling efficiency. Second， a sliding window combined with a coefficient of variation threshold criterion is constructed to accurately extract steady-state data from test run data， solving the problem of confusion between steady-state and transient data. Third， based on relative degradation analysis， the occurrence time of maintenance activities in the data is identified， and a segmented modeling strategy for performance degradation is established to effectively eliminate the interference of maintenance activities on degradation modeling. Finally， the random forest algorithm is used to fit the nonlinear multi-coupling relationship between engine gas path parameters， so as to achieve accurate prediction of turboshaft engine gas path performance. To verify the effectiveness of the proposed method， three segments of test run data are used for repeated verification. The results show that the average relative error of each prediction index is less than 1.09%. This method can effectively realize the accurate prediction of the degradation trend of turboshaft engine gas path performance， and provide reliable technical support for engineering practices such as engine fault diagnosis， health condition monitoring and performance prediction.