This　paper　presented　a　frequent　pattern　mining　algorithm　SSDTreeMiner　used　in　semi　structured　data　streams.　The　algorithm　gets　the　frequent　pattern　in　the　target　time　by　using　the　adding　and　deleting　principle　and　maintaining　the　sliding　windows,　at　the　same　time,　uses　the　time　attenuation　model　to　eliminate　the　influence　of　historical　data　in　the　mining　process,　compared　with　the　traditional　algorithms,　the　SSDTreeMiner　algorithm　can　deal　with　high　speed　and　complex　data　streams　and　complete　mining　tasks　real-timely　and　efficiently.