Machine learning (ML)-enabled high-throughput screening to predict potential electrocatalysts for the CO2 reduction reaction (CO2RR) offers new insights for energy conversion and environmental remediation. In this work, for the first time, we established a comprehensive electrocatalytic database containing ≈400 entries of CO2RR catalysts. Through decision tree analysis, correlation heatmaps, and feature importance ranking, we systematically decoded structure-property relationships. Among the tested algorithms, the nonlinear tree-ensemble method Random Forest Regression demonstrated superior predictive performance for CO2RR systems. Subsequent screening of 500 000 catalyst configurations generated by the the sequential model-based algorithm configuration method, using Expected Improvement as the evaluation metric, identified promising multinary alloy catalysts for C1 molecule production. Notably, BiSb-based alloys emerged as high-potential candidates for CO2RR applications. This ML-driven paradigm highlights the growing significance of artificial intelligence in materials discovery, synergistically combining screening efficiency, prediction accuracy, and proficiency in big data processing.