Based on the 47 “grain-to-feed” policies issued by the central and local governments in the country from 2015 to 2019. The text mining analysis methods have been used to extract high -frequency keywords, and then the “grain reform” policy evaluation index system with reference to existing research and policy characteristics was comprehensively constructed. A PMC index model was also built, and the randomly extracted 6 sample policy policies were quantified. The results showed that the average PMC index of each policy was 7.25, and 4 policies were rated as excellent and 2 were acceptable, suggesting that the overall situation was excellent. Among the two acceptable policies, although P2 was lower than the average in terms of policy areas, policy tools, policy audiences, policy evaluations and policy effects, it still served as a reference to a certain extent because it was released earlier than the national guidance documents. The policy content of P6 was relatively simple, leading to its score was lower than other policies. The common problem of the existing policies was that the time limit was short-term and temporary, and the policy of medium and long-term time limit was blank, which restricted the function of "grain to feed" and related industries to some extent. Finally, in view of the common and individual problems of various policies, optimization suggestions were also put forward.