For　the　problem　of　predicting　faults　in　the　status　of　a　shaft　furnace,　the　missed　alarm　rate　and　false　alarm　rate　have　not　been　improved　significantly　by　the　traditional　case-based　reasoning　(CBR)　method.　To　predict　faults　more　accurately,　an　improved　CBR-based　fault　prediction　method　(ICBRP)　is　proposed　in　this　paper.　This　ICBRP　is　composed　of　a　water-filling　theory-based　weight　allocation　(WFA)　model　and　a　group　decision-making-based　revision　(GDMR)　model.　According　to　the　optimal　allocation　mechanism　of　channel　power,　a　Lagrange　function　is　designed　to　calculate　the　weights.　Moreover,　the　credibility　of　historical　results　is　used　to　revise　the　predicted　results　via　the　definition　of　a　group　utility　function.　Then,　the　proposed　reasoning　strategy　can　obtain　more　reasonable　weights　and　take　full　advantage　of　comprehensive　information　from　the　retrieval　results.　Finally,　the　application　results　indicate　that　the　proposed　method　is　superior　to　traditional　CBR　and　other　methods.　This　proposed　ICBRP　significantly　reduces　the　missed　alarm　rate　and　the　false　alarm　rate　of　failure　in　the　furnace　status.　(C)　2013　Elsevier　Inc.　All　rights　reserved.