Rolling　bearing　fault　diagnosis　is　an　important　and　time　sensitive　task,　to　ensure　the　normal　operation　of　rotating　machinery.　This　paper　proposes　a　fault　diagnosis　for　rolling　bearings,　based　on　Generalized　Refined　Composite　Multiscale　Sample　Entropy　(GRCMSE),　Supervised　Isometric　Mapping　(S-Isomap)　and　Grasshopper　Optimization　Algorithm　based　Support　Vector　Machine　(GOA-SVM).　First,　GRCMSE　is　utilized　to　characterize　the　complexity　of　vibration　signals,　at　different　scales.　Furthermore,　an　effective　manifold　learning　algorithm,　named　S-Isomap,　is　applied,　to　compress　the　high-dimensional　feature　set　into　a　low-dimensional　space.　Subsequently,　GOA-SVM　classifier　is　proposed　for　pattern　recognition,　having　higher　recognition　accuracy　than　other　classifiers.　The　performance　of　the　proposed　method　has　been　verified　by　its　successful　application　in　a　rolling　bearing　fault　diagnosis　experiment.　Compared　with　the　existing　methods,　this　approach　improves　the　classification　accuracy　to　100%.　The　produced　results　indicate　that　the　proposed　method　can　effectively　detect　bearing　faults,　maintaining　high　accuracy.　©　2020　Elsevier　Ltd