In　this　paper,　we　use　the　wavelet　transform　and　the　ANN　model　to　predict　the　dynamic　traffic　demands.　Moreover,　we　use　ANN　models　to　estimate　the　task　performance　in　an　MEC　system.　By　leveraging　the　ANN　predictor　and　the　ANN　estimators,　we　present　a　resource　partitioning　scheme　for　data　transfers　in　the　MEC　system　to　dynamically　adjust　the　resource　partitioning　based　on　the　predicted　demands　and　the　performance　requirements.　Results　show　that　our　proposed　scheme　offers　near-optimal　results　compared　with　the　existing　scheme.　©　2022　IEEE.