The main purpose of this paper is to use machine learning algorithm to establish different classification models to evaluate and predict personal credit risk. In this paper, we take the data of give me some credit in the kaggle competition as an example. We take the seriousdlqin2yrs variable which is overdue for more than 90 days or worse as the target variable, and take other characteristic variables of the data as independent variables for modeling and analysis. In the stage of data preprocessing, this paper first uses the k-nearest neighbor method to fill in the missing values in the data, then processes the outliers in the data, and tests the multicollinearity among variables. In the process of model construction, logistic regression model is constructed by step-by-step screening method, decision tree model is constructed by cart algorithm, classification prediction model is constructed by SVM algorithm, and integrated model is constructed based on three algorithms. With the help of AUC value and ROC curve, by comparing the prediction effect of different models in training data set and test data set, it is found that the integrated learning model performs better, has higher classification effect and has stability.