It　is　important　to　select　similar　samples　in　the　local　modeling　for　fermentation　processes.　However,　previous　methods　didn＇t　consider　the　sample　weights　in　selecting　similar　samples,　especially　the　method　based　on　dynamic　time　warping.　To　improve　the　performance　of　the　local　model　through　selecting　the　more　similar　samples,　a　sample　similarity　measurement　method　based　on　weighted　Euclidean　distance　is　presented.　Based　on　the　analysis　of　the　affinity,　the　affinity　is　introduced　to　the　weighed　Euclidean　distance,　which　is　converted　to　a　similarity　measure　function,　and　the　function　is　used　to　the　similar　sample　selection　in　the　E.　coli　fermentation　process.　Combined　with　local　support　vector　machine　online　prediction　model　of　fermentation　product　is　established.　Experimental　results　show　that　compared　with　the　support　vector　machine　models　based　on　other　similarity　measure　function,　the　model　not　only　exhibits　a　higher　prediction　accuracy　and　a　better　generalization　ability,　but　also　the　prediction　time　is　significantly　reduced.　©,　2015,　Science　Press.　All　right　reserved.