The　rapid　progress　of　Internet　technology　brings　new　opportunities　for　the　development　of　science　and　technology,　and　the　emerging　network　video　technology　is　gradually　penetrated　into　people＇s　daily　life.　Due　to　the　rapid　development,　the　plentiful　video　contents　also　make　everyone　dazzling,　at　the　same　time,　the　video　users　with　multiple　geometric　growth　also　make　the　network　video　operators　not　know　what　to　do,　and　the　fundamental　technology　to　solve　this　problem　is　video　recommendation　technology.　This　paper　presents　an　algorithm　combined　with　the　neighborhood　latent　semantic　model,　the　new　model　retains　the　characteristics　of　recommended　explanation　in　neighborhood　algorithm,　and　expands　based　on　implicit　feedback　information　of　users,　which　has　further　improved　the　recommendation　efficiency.　The　new　model　adopts　the　field　method　of　User-CF,　this　paper　will　compare　the　operational　effects　of　basis　latent　semantic　and　the　latent　semantic　fusing　User-CF　neighborhood　mode　to　achieve　the　purpose　of　simulation.