Online　safety　monitoring　is　the　key　technology　to　realize　the　safe　operation　of　the　automatic　train　protection　(ATP)　system.　So,　based　on　the　probabilistic　model　checking　and　least　square　support　vector　machine　(LSSVM)　algorithms,　an　intelligent　quantitative　safety　monitoring　method　is　proposed　to　monitor　the　operational　safety　of　ATP　online.　To　begin　with,　the　dynamic　fault　tree　(DFT)　model　and　continuous-time　Markov　Chains　(CTMC)　model　of　ATP　are　established　based　on　the　fault-tolerant　structure　of　ATP.　Then,　the　reliability　and　safety　performance　of　DFT　are　evaluated　by　the　hierarchical　iterative　evaluation　method　when　considering　the　imperfect　fault　characteristics　of　the　critical　sub-equipment.　Furthermore,　continuous　stochastic　logic　(CSL)　is　introduced　to　represent　the　temporal　quantitative　safety　property.　For　the　defined　CSL　property,　the　CTMC　model　will　be　verified　by　probabilistic　model　checking　off-line,　and　the　verification　data　set　will　be　obtained.　The　distribution　regularities　of　maximum　reachable　probability　about　the　failure　rate　parameters　of　sub-equipment　will　be　achieved　by　training　the　obtained　verification　data　set　with　the　LSSVM　model.　Finally,　the　quantitative　safety　boundaries　(QSBs)　of　the　corresponding　quantitative　safety　levels　are　computed　by　the　designed　algorithm.　The　obtained　QSBs　can　be　used　for　monitoring　the　operational　status　of　ATP　online.