The invention relates to the field of gene detection technologies and biomedicine, in particular to a breast cancer patient recurrence risk 20 gene prediction model based on breast cancer single-celltranscriptome sequencing analysis and an establishment method and application thereof. The model is composed of 20 genes of CEBPD, SERPINA1, CD24, ERRFI1, BCL3, DSTN, BTG2, SERTAD1, SPINT1, BAMBI, LIMCH1, NFIA, SKP1, DHRS7, ODF3B, KRT7, ZFP36, CEBPB, BHLHE40 and UGDH. The breast cancer patient recurrence risk 20 gene prediction model provided by the invention provides more accurate judgment for long-term prognosis of a breast cancer patient, and provides a basis for selection of a postoperative adjuvant therapy scheme of the patient, so that individualized accurate treatment is realized, and meanwhile, a new research perspective is provided for research of breast cancer tumor stem cells.