Herein,　we　developed　a　flexible,　low-cost　non-enzymatic　sweat　sensing　chip　for　in　situ　acquisition　of　bioinformation　in　sweat　of　individuals　under　exercise　conditions　to　advance　personal　health　monitoring　and　medication　management　for　patients　with　Parkinson＇s　disease.　This　low-cost,　flexible,　wearable　sweat　sensor　consists　of　a　printed　screen　electrode　modified　with　g-C3N4　material　and　an　external　MSME　element.　The　doping　strategy　and　surface　activation　strategy　of　the　g-C3N4-based　exhibited　efficient　glucose　oxidase-like　activity　and　electrochemical　activity　when　testing　L-dopa　and　glucose　in　sweat.　The　optimized　signal　was　transmitted　to　a　smartphone　for　processing　12　individuals　with　simulated　dosing,　enabling　continuous　monitoring　of　L-dopa　metabolism　in　sweat　and　management　of　dosing.　The　generalization　ability　and　robustness　of　models　constructed　by　methods　such　as　multiple　linear　regression,　artificial　neural　networks,　and　convolutional　neural　networks　were　compared　cross-sectionally.　Deep　learning　models　based　on　artificial　neural　networks　help　develop　a　user-personalized　medication　administration　reminder　system,　which　provides　a　promising　paradigm　for　reliable　medication　supervision　for　Parkinson＇s　patients　in　the　Internet　of　Things　era.