In　data　mining　processing,　filling　incomplete　data　is　very　important　in　preprocessing　of　data　mining　project.　An　algorithm　based　on　time　series　similarity　measurement　for　incomplete　data　is　proposed,　this　approach　can　fill　missing　data　via　following　internal　rule　of　data　set　reasonably.　By　experiments,　we　can　draw　a　conclusion　that　algorithm　is　effective　on　condition　that　incomplete　data　are　no　more　than　half　of　the　whole　data　in　one　case.