The invention relates to an unbalanced
peptide hormone prediction method based on a
protein language model. The method solves the problem that in the prior art, the identification performance of
a peptide hormone prediction method is reduced under the condition of
class imbalance. The method comprises the following steps: S1, collecting and preprocessing a
data set; s2, embedding and coding sequence features; s3, constructing a model; s4, performing model performance optimization; s5, performing model performance evaluation; s6, performing practical application and expansion; and S7, realizing and deploying the
system. The method has the advantages that the potential
structure and function association of the
peptide sequence can be automatically learned under the condition that artificial
feature design is not needed. Meanwhile, aiming at the problem of
class imbalance that the number of
peptide hormone samples is far less than that of non-hormone peptides, the recognition capability of the model on
minority class samples is enhanced, and the balance and robustness of overall prediction are remarkably improved.