The simultaneous assessment of lift-off distance and thickness of metallic plates is significant in pulse eddy current testing. Pulsed eddy current testing technology holds the potential to assess wall thickness variations and local defects in metallic components through surface coatings. Firstly, a regression model based on Scikit-learn is proposed. The dimension of the full-waveform pulse eddy current assessment voltage signal is reduced by principal component analysis (PCA). A portion of the reduced data is used to train the regression model, while another portion is used to validate the model’s effectiveness. Secondly, pulse eddy current testing signals with different lift-off distances and metal plate thicknesses are obtained through analytical calculations for analysis. Finally, the robustness of the noise method is investigated by adding Gaussian noise to the analyzed signal. The test results show that the regression method based on Scikit-learn proposed could be used to assess the lift-off distance and metal plate thickness simultaneously from the full-waveform pulse eddy current testing voltage signal, and it has good robustness to noise.