Fraud user identification algorithm framework fusing user burying point behavior information

The invention discloses a fraudulent user identification algorithm framework fusing user burying point behavior information, which comprises the following steps of: (1) recording a click behavior and a stay time length of a client within one hour before a user applies for credit granting, and generating a ternary combination of a page (p), a click button (b) and the stay time length (t); compared with a traditional RFM feature derivation mode, the method has the advantages that current popular word vector representation is adopted, and sequence behaviors can be better represented while information loss is reduced; a random walk algorithm is utilized to solve the problem of insufficient short-term sample size, so that more random sequence behaviors are generated to be used for training word vectors; for more concerned time interval characteristics in the sequence behavior information, the combination of the behavior information is converted into page + button + stay duration classification from page + button +