The invention discloses a customer classification method and device based on cost sensitivity and semi-supervised classification, and the method comprises the steps: obtaining a data set L with a class label, a data set U without class label, and a test set Test; employing a random subspace method for the data set L with the class label and the data set U without class label, and training N basicclassification models CS; employing the N basic classification models CS for the classification of the samples in the test set Test, and obtaining N intermediate classification results R1, R2,..., RN;carrying out the majority voting integration of N intermediate classification results R1, R2,..., RN, and obtaining a final classification result. The method combines cost sensitivity learning, semi-supervised classification and random subspace, can achieve the processing of data of the imbalanced types in a better way through the cost sensitivity learning, also can achieve the utilization of a large amount of information in the samples without the class label through semi-supervised learning, also can improve the target customer selection performance of a model through the random subspace, and obtains the better target customer selection performance.