Historical　cases　have　been　accumulated　to　use　the　problem-solving　experiences　of　similar　types　of　historical　cases　to　solve　current　problems.　The　existing　studies　has　shown　its　potential　to　generate　alternative　accurately　but　also　exposed　some　challenges,　such　as　the　precision　of　case　information　representation,　the　accuracy　of　case　match,　and　the　validity　of　the　generated　alternative.　This　paper　proposed　a　case-driven　emergency　decision-making　model　(EDM)　based　on　probabilistic　linguistic　bidirectional　projection　to　overcome　these　challenges.　A　new　attribute　match　measurement　based　on　probabilistic　linguistic　bidirectional　projection　under　probabilistic　linguistic　term　sets　(PLTSs)　environment　not　only　preserve　the　integrity　of　the　case　information　but　also　facilitate　valid　retrieval　results　and　guarantee　an　interpretable　decision-making　process.　An　aggregation　model　based　on　the　best–worst　method,　and　a　distance-based　optimization　model　considering　the　interactions　between　attribute　combinations　is　then　constructed　to　weigh　all　the　attributes　to　determine　similarities　between　the　current　and　historical　cases,　which　can　make　the　match　result　more　reasonable.　Afterwards,　a　bidirectional　projection–gained　and　lost　dominance　score　(GLDS)　method　is　developed　and　applied　to　improve　the　fairness　and　validity　of　the　retrieval　results.　Finally,　a　case　study　of　suspected　cases　of　Coronavirus　Disease　2019　(COVID-19)　demonstrates　the　practicability　of　the　proposed　proposal,　and　comparative　analysis　illustrates　its　superiority　over　alternatives.　©　2023　Elsevier　Ltd