The　research　of　this　topic　is　based　on　financial　risk　control,　and　Sunshine　Jinke＇s　risk　prevention　and　control　is　taken　as　a　specific　research　case.　The　data　set　uses　user　consumption　data　used　by　Sunshine　Jinke.　The　data　source　is　a　bank＇s　customer　consumer　loan　records　in　the　past　five　years.　The　feature　fields　are　extracted　according　to　the　degree　of　correlation　between　the　data　and　the　repayment　rate,　and　combined　with　the　convolutional　neural　network　and　the　recurrent　neural　network,　these　feature　fields　are　processed　for　the　differentiation　of　credit　data　and　fraud　data,　and　finally　matched　with　user　information　and　scored,　matching　The　process　uses　the　singular　matrix　factorization　idea.　Through　experimental　papers,　this　idea　has　good　stability　and　accuracy　in　the　field　of　risk　control　forecasting.　©　2021　IEEE.