In　the　large　for　gestational　age　infant’s　classification　and　prediction,　noisy　features　are　distilled　to　improve　the　classifier　performance.　It　is　accomplished　with　the　creation　of　a　suitable　feature　vector　followed　by　GridSearch-based　Recursive　Feature　Elimination　with　Cross-Validation　(RFECV)　scheme.　It　attempts　to　elect　features　that　are　influential　and　independent.　We　executed　experiments　on　the　data　obtained　from　the　National　Pregnancy　and　Examination　Program　of　China　(2010–2013).　The　results　are　compared　with　the　results　already　reported　in　the　literature.　The　GridSearch-based　RFECV　scheme　exhibited　smaller　features　subset　size　with　an　increased　classifier　performance.　The　precision　and　area　under　the　curve　(AUC)　scores　are　drastically　improved　from　0.7134　and　0.7074　to　0.96　to　0.86　respectively.　Therefore,　pediatricians　are　suggested　to　use　fifty-three　features　subset,　ranked　by　GridSearch-based　RFECV　scheme　using　Support　Vector　Machine　(SVM)　for　the　establishment　of　an　efficient　LGA　prognosis　process.　©　2020,　Springer　Nature　Singapore　Pte　Ltd.