the　paper　puts　forward　a　new　classification　method　of　association　rules　that　is　based　on　support-significant　structure.　Starting　from　the　characteristics　of　customer　segmentation,　this　method　introduces　a　up-to-date　rule　evaluation　index-significance　during　the　produce　process　of　classification　rules,　therefore,　the　evaluation　and　selection　of　principle　have　serious　statistical　basis.　After　simple　comparison　of　the　real　example,　we　prove　the　comparative　superiority　of　introducing　significance　into　the　customer　segmentation.