The　traditional　collaborative　filtering　algorithm　only　pays　attention　to　the　rating　by　users.　In　reality,　however,　user　and　item　information　is　always　changing　with　time　flying.　Therefore,　recommendation　systems　need　to　take　time-varying　changes　into　consideration.　The　collaborative　filtering　algorithm　which　is　based　on　Forgetting　Curve　and　Long　Tail　theory　(FCLT)　is　introduced　for　the　above　problems.　The　following　two　points　are　discussed　depending　on　the　problem:　First,　the　user-item　rating　matrix　can　update　in　real　time　by　forgetting　curve;　secondly,　according　to　the　Long　Tail　theory　and　item　popularity,　a　further　similarity　calculation　method　is　obtained.　The　experimental　results　demonstrated　that　the　proposed　algorithm　can　effectively　improve　the　recommendation　accuracy　and　alleviate　the　Long　Tail　effect.