Personalized　recommendation　in　Internet　plays　a　very　important　role,　but　it　suffers　from　the　problem　of　how　to　capture　users＇　feedback　and　predict　their　interest.　Considering　reading　activities　can　be　a　good　indicators　to　user　interest,　in　this　paper,　we　predict　the　users＇　interest　based　on　their　reading　behaviors　analysis.　First,　the　users＇　reading　activities　are　captured　and　modeled　with　hidden　Markov　Model　(HMM).　Then　by　using　these　specific　interest　models,　we　can　predict　users＇　interest　by　analyzing　reading　activities.　Last,　a　RSS　reader　client　is　designed　based　on　above　methods.　Experiments　results　in　10　users　with　more　than　1500　Web　pages　show　that　the　user　interest　can　be　predicted　with　an　acceptable　accuracy　with　our　method.