More　and　more　sensor-based　information　systems　have　been　put　into　use,　and　the　data　from　the　sensors　are　growing　rapidly.　The　emergence　of　multi-sensor　data　fusion　provides　more　mining　methods　and　can　solve　the　problem　from　many　different　viewpoints.　This　paper　presents　an　improved　outlier　detection　method　for　the　building　strain　monitoring　data　by　using　multi-sensor　data　fusion.　According　to　the　real-time　temperature　data,　the　strain　data　at　the　same　period　can　be　segmented.　In　each　segment　the　traditional　outlier　detection　is　used.　Comparing　the　segment　with　the　non-segment　method,　it　is　proved　that　the　method　using　multi-sensor　data　fusion　improves　the　efficiency　of　outlier　detection.　©,　2015,　Binary　Information　Press.　All　right　reserved.