High accuracy optical clock is regarded as an important tool for testing general relativity, relativistic geodesy and dark matter detection. So far ytterbium optical clock shows the best stability, which reaches 10－19 level after averaging 36 h. Control unit is the vital part of the ytterbium optical clock. It mainly takes on time sequence generation for the clock system. In this paper, a field programmable gate array (FPGA) control unit based on Python language is introduced to ytterbium optical clock. On one hand, it can generate various control signal for ytterbium optical clock by compiling and calling the subroutine. On the other hand, it can be used for data storage and processing during the clock operation. The results show that data processing with the FPGA system can be perfectly matched with traditional method dealt with Matlab and Origin program, and cold atoms with micro Kelvin level and clock transition spectra with Hertz level are obtained in ytterbium clock system. More importantly, based on the noise analysis database in earlier experiment, fitting to the complex clock transition spectra with multi-Gaussian superimposition function can act as a method for machine learning, which is a better way for experimental process analysis and fault diagnose. It would benefit for gravitational wave detection and hunting for dark matter with more accurate atomic optical clock based on machine learning in the future.