Urban　subway　is　taken　by　people　in　different　frequencies,　thus　leading　them　to　present　different　dependency　varieties　on　this　mode.　Yet,　how　those　passengers　who　possess　low　dependency　on　urban　subway　travel　is　less　investigated.　Under　this　background,　we　propose　a　framework　to　uncover　passengers＇　dependency　variety　on　stations＇　functions　in　urban　subway.　To　begin　with,　nine　features　regarding　travel　frequency　and　time　are　extracted　from　100　million　transaction　records　generated　by　11.45　million　passengers　in　a　month.　Thus,　their　travel　dependency　on　urban　subway　is　quantified.　These　features　are　clustered　into　5　distinct　levels　via　the　k-means　algorithm,　before　an　inference　of　subway　stations＇　functions　from　236,040　POI　data　sources　via　the　LDA　approach.　In　this　way,　passengers＇　travel　purposes　can　be　identified.　How　passengers　with　different　dependency　levels　behave　in　subway　stations　in　space　and　time　is　further　explored　in　a　visualization　way.　The　intuitive　experimental　results,　validated　by　priori　user　experiences　and　land-use　plan　of　Beijing,　show　that　among　the　5　levels　of　dependency　varieties,　passengers　in　the　first　two　groups　present　a　relatively　strong　dependency　on　urban　subway.　Meanwhile,　passengers　in　the　rest　three　groups　possess　a　low　dependency　on　urban　subway　and　display　extreme　travel　patterns　in　time　and　frequency,　greatly　increasing　management　difficulties　for　transit　operators.　Findings　in　this　research　help　distinguish　passengers　with　low　levels　of　subway　dependency　and　grasp　how　those　passengers　without　striking　dependency　travel　by　subway　and　what　for　so　that　practitioners　can　conduct　an　accurate　risk　assessment　on　them.