In　this　article,　we　study　a　general　single-index　model　with　diverging　number　of　predictors　by　using　the　adaptive　Elastic-Net　inverse　regression　method.　The　proposed　method　not　only　can　estimate　the　direction　of　index　and　select　important　variables　simultaneously,　but　also　can　avoid　to　estimate　the　unknown　link　function　through　nonparametric　method.　Under　some　regularity　conditions,　we　show　that　the　proposed　estimators　enjoy　the　so-called　oracle　property.