Numerous　educational　institutions　utilize　data　mining　techniques　to　manage　student　records,　particularly　those　related　to　academic　achievements,　which　are　essential　in　improving　learning　experiences　and　overall　outcomes.　Educational　data　mining　(EDM)　is　a　thriving　research　field　that　employs　data　mining　and　machine　learning　methods　to　extract　valuable　insights　from　educational　databases,　primarily　focused　on　predicting　students＇　academic　performance.　This　study　proposes　a　novel　federated　learning　(FL)　standard　that　ensures　the　confidentiality　of　the　dataset　and　allows　for　the　prediction　of　student　grades,　categorized　into　four　levels:　low,　good,　average,　and　drop.　Optimized　features　are　incorporated　into　the　training　process　to　enhance　model　precision.　This　study　evaluates　the　optimized　dataset　using　five　machine　learning　(ML)　algorithms,　namely　support　vector　machine　(SVM),　decision　tree,　Naive　Bayes,　K-nearest　neighbors,　and　the　proposed　federated　learning　model.　The　models＇　performance　is　assessed　regarding　accuracy,　precision,　recall,　and　F1-score,　followed　by　a　comprehensive　comparative　analysis.　The　results　reveal　that　FL　and　SVM　outperform　the　alternative　models,　demonstrating　superior　predictive　performance　for　student　grade　classification.　This　study　showcases　the　potential　of　federated　learning　in　effectively　utilizing　educational　data　from　various　institutes　while　maintaining　data　privacy,　contributing　to　educational　data　mining　and　machine　learning　advancements　for　student　performance　prediction.