Observing　and　analyzing　travel　behavior　is　important,　requiring　understanding　detailed　individual　trip　chains.　Existing　studies　on　identifying　travel　modes　have　mainly　used　some　travel　features　based　on　GPS　and　survey　data　from　a　small　number　of　users.　However,　few　studies　have　focused　on　evaluating　the　effectiveness　of　these　models　on　large-scale　location　data.　This　paper　proposes　to　use　travel　location　data　from　an　Internet　company　and　travel　data　from　transport　department　to　identify　travel　modes.　A　multiple　binary　classification　model　based　on　data　fusion　is　used　to　find　out　the　relationship　between　travel　mode　and　different　features.　Firstly,　we　enlisted　volunteers　to　collect　travel　data　and　record　their　travel　trip　process　using　a　custom-developed　WeChat　program.　Secondly,　we　have　developed　three　binary　classification　models　to　explain　how　different　attributes　can　be　used　to　model　travel　mode.　Compared　with　one　multi-classification　model,　the　accuracy　of　our　model　improved　significantly,　with　prediction　accuracies　of　0.839,　0.899,　0.742,　0.799,　and　0.799　for　walk,　metro,　bike,　bus,　and　car,　respectively.　This　suggests　that　the　model　could　be　applied　not　only　in　engineering　practice　to　identify　the　trip　chain　from　Internet　location　data　but　also　in　decision　support　for　transportation　planners.