Crime　prediction　is　of　great　significance　to　the　food　safety　defense　work　of　major　events.　The　traditional　crime　prediction　depends　on　the　experience　of　police　officers,　which　is　highly　subjective　and　can　not　be　predicted　in　advance.　This　paper　analyzes　and　processes　food　safety　police　data,　combines　the　characteristics　of　food　crimes,　uses　stacking　model　fusion　method　to　predict　the　food　crime　tendency　of　key　personnel,　and　verifies　the　model　through　the　Recall　Rate.　The　results　show　that　the　integrated　learning　model　has　high　accuracy　and　can　effectively　predict　the　food　crime　tendency　of　key　personnel.　©　2021　IEEE.