The embodiment of the invention discloses a credit review model construction method, device and system, according to the scheme provided by the embodiment of the invention, the historical data is segmented and screened in advance;, some modeling individuals with high missing rate and some characteristics with relatively poor prediction capability are removed; the remaining data is taken as a feature candidate set; based on an ensemble learning model of Logistic Regression, the credit review model is constructed. According to the technical scheme, the timeliness and efficiency of credit reviewcan be greatly improved, the loan ability and repayment willingness of a borrower can be analyzed from all aspects and dimensions based on historical data, the malicious fraud phenomenon can be recognized, and the prediction accuracy and robustness are improved.