With　the　rapid　expansion　of　the　amount　of　information　in　social　media　such　as　various　news　and　information　software,　people　urgently　need　to　realize　the　automatic　classification　of　this　information　to　help　users　quickly　find　the　information　they　need　and　filter　spam.　Aiming　at　the　curse　of　feature　dimension　and　non-semantic　features　in　traditional　text　classification　models,　this　paper　studies　the　classification　of　article　text　based　on　the　Word2vec　model.　One　of　the　shortcomings　of　the　Word2vec　model　is　that　the　essentiality　of　words　in　diverse　texts　is　not　the　same,　so　this　paper　introduces　the　TFIDF　model,　which　can　weight　Word2vec　word　vectors　to　achieve　a　weighted　Word2vec　classification　model.　At　last,　the　weighted　Word2vec　and　TFIDF　models　are　combined.　©　2022　IEEE.