The output bandwidth of the arbitrary waveform generator (AWG) is limited by the bandwidth of the digital to analog converter (DAC). The frequency interleaved DAC (FI-DAC) could enhance bandwidth enhancement effectively. However, non-ideal characteristics of analog components cause peak amplitude frequency errors in the edge frequency bands of FI-DAC. These errors reduce the flatness of the output signal and degrade system performance. To address this issue, this article proposes an improved FI-DAC precalibrator that focuses on addressing the peak nonlinear amplitude frequency errors. Firstly, the linear phase error between the two channels of the FI-DAC is calibrated. Secondly, the algorithm utilizes support vector regression (SVR) to realize a precalibrator by formulating a regression model for the initial calibration of amplitude frequency errors. Thirdly, the algorithm integrates locally weighted learning (LWL) to assign adaptive weights to the edge frequency band. Finally, with a single-channel DAC sampling rate of 1.25 GSa/s, the application of the FI-DAC achieves an output bandwidth of 850 MHz, improving the signal output frequency range. Experimental results show that the minimum flatness within the passband is -0.061 dB, and the maximum flatness is 0.032 dB, which is close to the ideal flatness of 0 dB. Further validation is conducted on the 5 GSa/s experimental platform. Compared with other algorithms, the SVR-LWL precalibrator achieves higher accuracy in calibrating peak amplitude frequency errors in the FI-DAC.