Predicting　chaotic　time　series　by　fuzzy　inference　systems　is　one　of　active　research　areas.　This　paper　concerns　Mamdani　and　Sugeno　fuzzy　inference　systems　in　the　application　of　chaotic　time　series　prediction.　These　two　types　of　fuzzy　inference　systems　were　compared　through　four　sets　of　experiments　on　a　number　of　chaotic　time　series　to　evaluate　their　prediction　performance　in　terms　of　model　generalization,　execution　time,　structure　complexity,　and　noise　tolerance.　The　experimental　results　indicate　that　it　may　as　well　choose　Sugeno　fuzzy　inference　system　as　a　predictor　in　the　first　place　for　the　situations　where　the　training　data　are　not　severely　corrupted　by　noise.