Modeling　passengers＇　hidden　relation　proximity　is　beneficial　for　target　detection,　shared　parking,　or　carpooling.　Most　efforts　concentrated　on　modeling　individuals＇　relation　proximity　based　on　their　profile　similarities　or　interaction　activities,　without　considering　neighborhood-driven　relation　expansions　in　the　transportation　domain.　This　research　utilizes　smart　card　data　generated　by　transit　riders　in　Aug.　2015,　Beijing,　to　explore　their　relation　proximity　based　on　a　proposed　weighted　relation　strength　model,　which　measures　not　only　the　existing　causal　dependency　of　profile　similarity　and　interaction　activity,　but　also　the　neighborhood-driven　relation　strength　of　passengers.　To　begin　with,　by　knowing　that　cards　with　adjacent　IDs　may　be　held　by　correlated　passengers　who　possess　similar　temporal　probability　distribution　sequences　at　same　stations,　we　cluster　the　above　sequences　and　IDs,　respectively,　and　extract　4,162　pairs　of　passengers　to　build　a　customized　sample　dataset,　within　which,　456　pairs　are　confirmed　as　true　positives.　All　passengers＇　pairwise　similarities　are　characterized　by　their　profile　similarities,　interaction　activities,　and　neighborhood　interactions,　which　are　then　automatically　learned　by　the　proposed　model.　The　experimental　results　confirm　the　method＇s　robustness　and　precision　in　inferring　passengers＇　relation　proximity.