A Method to Examine the Effects of Blood Glucose on Human Health by Combining Machine Learning Algorithms

Abstract

Diabetes iis ia icommon idisease iand iits iearly isymptoms iare inot ivery inoticeable, iso ian iefficient imethod iof iprediction iwill ihelp ipatients imake ia iself-diagnosis. iHowever, ithe iconventional imethod ito iidentify idiabetes iis ito imake ia iblood iglucose itest iby idoctors iand ithe imedical iresource iis ilimited. iTherefore, imost ipatients icannot iget ithe idiagnosis iimmediately. iSince ithe iearly isymptoms iof idiabetes iare inot iobvious iand ithe irelationship ibetween isymptoms iand idiabetes iis icomplex. iThe iprocess iof iMachine iLearning iis ito itrain ia icomputational ialgorithm ifor iprediction ibased ion ia ibig idataset. iIt iis ipopular ifor iits iefficiency iand iaccuracy. iAlso, iit ihas ithe iadvantage iof idealing iwith itons iof idata, iso iwe ican imake idiagnoses ifor iplenty iof ipatients iin ia ishort itime iand ithe iresult iwill ibe imore iaccurate. iIn ithis istudy, iwe iused iclassical imachine ilearning imodels iKNN, ito imake ia iprediction imodel ifor idiabetes idiagnosis. iOur idata iwas ifrom iUCI iMachine iLearning iRepository. iWe iconduct iparameter ituning ion ieach imodel ito itrade-off ibetween ithe iaccuracy iand icomplexity. iThe iaccuracy iof iKNN iof ithe itest idataset iachieves i81 ipercent, iwhich iis ithe ibest imodel ifor ipredicting idiabetes.