The　Single-Pass　clustering　algorithm,　its　two　main　disadvantages　are　easily　affected　by　the　orders　of　inputs　of　text　and　low　precision　when　we　use　it　to　process　the　network　text　clustering.　Through　introducing　the　concept　of　seeds　of　topic,　the　paper　proposed　an　improved　Single-Pass　clustering　algorithm　which　inherited　the　main　means　of　Single-Pass　clustering　algorithm.　The　experiment　results　showed　that　the　improved　algorithm　could　not　only　improve　the　speed　of　clustering,　but　also　decrease　the　probabilities　of　miss　detection,　false　detection,　and　the　cost　of　wrong　detection.　The　improved　Single-Pass　clustering　algorithm　that　has　improved　the　quality　of　clustering　and　topic　detection　both　has　high　practicability　and　good　reference　value　to　the　research　of　analysis　for　internet　public　opinion.