2. INTRODUCTION TO MACHINE LEARNING
• SupportVectorMachines(SVM).
• RandomForests.
DuringtheseDifficultTimesHealthisutmostImportanceto Everyone.PeoplearemoreconsciousabouttheirHealthnow more than ever. The pandemic has exposed of how much havoccanbecausedbythingswhichwehavebeenneglected for years. Specially the attention was drawn to comorbid diseaseswhichplayedaimportantroleinthepandemic.The comorbid diseases can be termed as diseases which alter yourentirelifestyleandcanimpactyourfuturelifestyletoo. Inshortthesearetheoneswhichneedtobetakenseriously. Some of The Examples being Diabetes, Heart Disease, Chronic Kidney Disease, Liver Disease, Lung Failure etc. Moreover,thepeoplewhicharehavingthesediseaseswere given special priority when the covid vaccination drives were conducted. This just shows how important these diseasesarehowtheyshouldbetakenveryseriously.Now, the question arises that how can we detect or predict if a particularpersonhasthisdiseaseatpresentormayhaveit intheFuture.ThisiswherethedomainofMachineLearning and Data Science comes in. Now that we have sufficient introduction about comorbid diseases lets shift our focus towardstheintroductionofMachineLearninganditsrelated termswhichareextensivelyusednowadaysintheindustry.
Figure 1 SubDomainsofMachineLearning
Machine Learning or Popularly abbreviated as ML is a techniqueorprocessinwhichthemachinelearnsbyitself with the help of different computer algorithms which are trained and tested on data. It turns applies the same approachandalgorithmonthedataagainwithoutneeding any further human intervention. This is as simple the definitionofMachineLearningcanget.MLasadomainhas3 different Subparts to it namely, Supervised Learning, Unsupervised Learning and Re enforcement Learning. All theseSubdomainshavesomeAlgorithms whichfallunder them,tounderstanditbetterpleasereferthefigurebelow.
Supervised Machine Learning
Itisdefinedasthedomainwhereeverythinghappensina guidedpath,beinganalogoustobetaughtbyateacher. In this the algorithms are trained well on train data and are thentestedonthetrainingdata.Thissubdomainisusedfor theclassificationsandpredictionbasedonthedata.Someof the different algorithms which fall under supervised machinelearningare:
• LinearRegression.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page726 HealthOrzo – Your Health Matters Vaishali Jabade1 , Sarthak Bhake2 , Sahil Parekh3, Sakshi Kulkarni4, Sejal Sayam5, Shajjad Shaikh6 1Assistant Professor, Vishwakarma Institute of Technology, Pune, Maharashtra, India 2 6Students, Dept. of Electronics and Telecommunications, Vishwakarma Institute of Technology, Pune *** Abstract - Health is Wealth. This Sentence is becoming more algorithmsHealthOrzofriendlyAllfromAlgorithmsSupportdifferentthatdataandwondersArtificialpossible.moreandmoreaccuratenowadays.PeoplearebecomingmoreandhealthconsciousandwanttobeinthebestshapeOurProjectcaterstothesameneedofthePeople.IntelligenceandMachineLearninghavedoneinthehealthcareindustrybyusingtheclassificationregressionalgorithmstopredictdiseasesbasedonthegivenbytheuser.TheProjectPredicts4seriousdiseasesisDiabetes,Heart,KidneyandLiverAilmentusingsupervisedMachineLearningAlgorithmsLikeVectorMachines,LogisticRegressionetc.ThearetrainedandtestedondatasetswhicharetakenKaggleforthedifferentdiseasesthattheProjectPredicts.the4machinelearningmodelsareintegratedinauserwebsiteusingtheflasklibraryatthebackend.isthusawebsitehavingall4predictionat1singleplaceintheformofawebsite. Key Words: Machine Learning, Supervised Machine Learning, Support Vector Machine (SVM), Logistic Regression, Random Forest, Python, Flask, Health 1.INTRODUCTIONPrediction.
Asyoucanseethedistinctionbetweenthe3subdomains,let usunderstandwhichalldifferentAlgorithmsfallunderthese subdomains.
• LogisticRegression.
thealgorithmA.H.Chenet.al[3]hadimplementedRandom Forest Classifier and had an 80% accuracy but opted for logisticregressionasouralgorithmandgot83%accuracy. NowcomingtotheKidneyDiseasePredictionmodel,G.Chen etal[4]elaboratedontheparameterswhichareconsidered duringkidneydiseaseandpredictionandimplementedthe predictionusingneuralnetworks.Nowastheparameterare fixed, we fixed our dataset which was elaborated by P. Chittoraet.al[5].DiabeteswasalsoaverykeyModelinour Projectandthedatasetwasalreadydecidedbutwenowhad to choose the algorithm. A. Anand et.al [6] implemented randomforestclassifierbuttheaccuracyonourdatasetwas quitelowforthisparticularalgorithm.A.Miret.al[7]intheir researchshowedthatmostofthepeoplehaveoptedSupport Vector Machine (SVM) as their optimal algorithm for diabetesprediction,henceweoptedforthesamealgorithm. S. Sontakke et.al [8] in their paper illustrated about the classificationandpredictionofthemodel.Thus,wedecided randomforestasouralgorithmforourlivermodel.
4. METHODOLOGY
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page727 • DecisionTrees. • NaïveBayes.etc. Unsupervised Machine Learning As you must have probably guessed by its name, the algorithmswhichfallunderthisdomainareperformedon datathatisnotsupervisedorlabeled.Themodelswhichare builtusingunsupervisedalgorithmsareself sufficienttofind patternsandinsightsfromthedataontheirown.Thisbeing analogoustohowourhumanbrainlearnsnewthingsand findspatterns/drawsinsightsformagivenparticularthing. Some of the different algorithms which fall under unsupervisedmachinelearningare: • K meansClustering. • HierarchicalClustering. • K nearestNeighbors(Knn).Etc. Figure 2 Differencebetweenthe3subdomainsina visualformat 3. RELATED WORK During the research of this Project, we scrapped many researchpapersinordertostudydifferentapproachesand algorithmswhichwouldsuitourproblemstatementinthe best way. As mentioned earlier we are going to predict 4 differentcomorbiddiseaseslikediabetes,Heart,kidneyand liverdisease.Inordertobuild4differentMachineLearning Models each catering to a particular disease we had to go through different papers in order to choose the optimal algorithms with the best accuracy and the datasets which suit those particular algorithms. While doing Literature survey for heart disease prediction we had to lock upon whicharethekeyparameterswhichweneedtoconsiderfor the prediction. A. Gavhane et.al [1] recently reported and mentionedthedifferentfeaturesnecessaryforheartdisease prediction,whichbeingBodyMassIndex,HeartRate,Fasting Blood Sugar, Age, Cholesterol etc. After locking in on the parametersrequired,wefinalizedthedatasetwehaveused forheartpredictioncourtesyofA.Singhet.al[2]wherethey had specified the dataset and elaborated on each feature. Nowthedatasetbeingfinalizedforheartwehadtochoose
ThiswasallfortheLiteratureSurveynowlets’havealook andwhatisthemethodologyanddifferentstepswhichare involvedinthebuildingofthese4MachineLearningModels.
While Building Machine Learning Models there some steps/methods which are carried out every time. You can alsocallthemasSOPStandardOperatingProcedurewhile buildingMachineLearningModels.Let’shaveaLookatallof themonebyone.
• Seaborn • Sci kitLearn
ThereareDifferentTypesofPlotswhichareuseful,someof themare.
Thisisaverykeysteptowardstobuildingofourmodel.This isthe step where all our data getscleanedand visualized. Afterthedataisvisualizedusingdifferentchartsandmetrics it is easy to draw insights from charts that are presented.
• Numpy • Pandas • Matplotlib
B. Data Pre processing and Visualization
A. Importing the Libraries First of all, we need to import the necessary libraries and toolsneededforourproject.Beitsomepre processingtools or ML Enabling Libraries. These are the Following Libraries/Frameworks we have used/imported for our Project.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page728 • Histogram. • ScatterPlot. • BoxPlot. • HeatMaps. • PieCharts. • DistributionPlot. Figure 3 DifferentTypesofChartsandVisualization Techniques
Figure 6 DatasetusedforDiabetesPrediction
Figure 5 Savingourmodelinapicklemodel.
Figure 7 SupportVectorMachineRepresentation
A.1 Algorithm To predict the diabetes, we have used Support Vector Machine(SVM)Algorithm.ItisanAlgorithmwhichcomes under supervised machine learning and this algorithm is mainly used for classification purposes. The main Goal of SVMistocreateabestlineorsomewhatlikeaboundarythat cansegregatendimensionalPlaceintoclassessothatwecan segregate future data into the correct category. The boundary which segregates the data points is called Hyperplane.
Figure 4 TrainTestSplitusingSci kitLearn
D. Saving the Model Afterwehavesuccessfullycompletedbuildingourmodelit’s timetosaveitintheformofafileforfurtheruse.Tosave theseparticularmodelswehaveusedpicklelibrarywhich enablesswiftsavingofmodelsandtheloadingitanywhere asfilewitha.pklextension.Wehavetakenthefulladvantage of this property while developing our flask app where we renderaparticularmodel basedontheinputgivenbythe user.
C. Train Test Split
This is step where we split our dependent variables and independentvariableintotwodifferentseriesordataframes andthenweperformthetraintestsplitwherethedividethe data set into four variables namely, X_test, y_train, X_test, y_test . We import the train test split call from a library named sk learn which has in built instances for this particularsplit.Youcanreferthesyntaxgiveninthefigure attachedbelow.
5. IMPLEMENTATION
OurProjecthasbeendividedinto2parts,oneisthemodel folderandotheristhewebpartwherewehavetheHTML TemplatesandtheFlaskapprunningintheBackendofthe code.Wehaveused4Modelsnamelytocatertoourdiseases thatisHeart,Diabetes,LiverandKidney.We’llFocusoneby oneonallourmodels.
A. Diabetes Model FortheDiabetesModelwehaveimportedthedatasetwhich hasthefollowingfeaturesattachedintheFigurebelow.
In order to predict this particular disease, we have used Logistic Regression as our Algorithm. This Particular Algorithm predicts the output of the dependent variable. Thus,theOutputshouldalsobeadiscretevalue.Itreturnsa probabilistic value which has a value between 0 to 1. The mathematical representation of logistic regression can be foundinthefigureattachedbelow.
C. Liver Model
Figure 10 LogisticRegressionRepresentation
B.2 Accuracy Score
B. Heart Model
Figure 11 AccuracyScoreofHeartModel
A.2 Accuracy Score
TheliverModeliswherewefirstclassifyandthenpredict thedisease basedonthedatasetwhichis beingused. The datasethasdifferentfeaturesandpigmentratioswhichare keyinassessingwhetherapersonhasafitliverornot.These Ratios not only signify one’s liver’s well being but also signify the overall health of a person. The features can be seenfromtheimageattachedbelow.
Figure 12 DatasetUsedforLiverDiseasePrediction C.1 Algorithm
The score which represents the performance of the algorithm is called the Accuracy score. We have Achieved 75%AccuracyonthismodelbyusingSVM.
Figure 9 DatasetusedforHeartDiseasePrediction
Figure 13 RandomForestClassifier
The dataset which is used for this model has various differentfeatureswhichareinfluentialintheprediction.The featurescanberecalledfromthefigureattachedbelow.
The Accuracy Score of this model after using Logistic Regression81%.Theaccuracytestingisdoneonbothtrain dataandtestdata.
Figure 8 AccuracyScoreofDiabetesModel
B.1 Algorithm
ForthisPredictionwehaveusedRandomForestclassifier and regressor. This particular algorithm comes under ensemble learning. The Algorithm uses multiple decision tress to classify and predict the outcome of a particular situation.ThenominalWorkingofRandomForestscanbe seenintheFigurebelow
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page729
Thelivermodelafterapplyingtherandomforestalgorithm givesanaccuracyof73%onthetestingdataset.
Figure 16 ConfusionMatrixandAccuracyScoreof KidneyModel
C.2 Accuracy Score
Thekidneymodelanddatasetareonemostcomplicatedyet the most interesting case. The dataset has 26 different columns containing various features used in the Machine Learning Model. The dataset can be seen in the figure attachedbelow.
TheAlgorithmusedforthisissameastheLivermodelthatis RandomForestClassifierandRegressor.Thetheoryissame asmentionedearlierinthepaper.
Figure 15 DatasetusedforKidneyDiseasePrediction
In the Figure Above it is the function which accepts the HTML Form input from the respective templates and typecastsitintoalistandchangesitsdatatypetofloat.After thisisdonethevariableinwhichtheinputisstoredisthen passed to the function which has the Machine Learning Modelloaded.(ThefunctionshowedinFigure17)
D. Kidney Model
Figure 17 App.pyfile
D.1 Algorithm
TheresultsoftheML modelsarerenderedona front end userinterface.InourProjectwehavecreatedasimplestatic websitewhichwillserveasapresentableinterfaceforour project. The website contains a landing page and 4 other pagescateringtodiseasesDiabetes,Heart,KidneyandLiver respectively.The4pagescontainhtmlformswhichtakethe inputfromtheuserandpassthoseinputstothemodelswith the help of Flask app which is named as app.py. The prediction is basically in 2 categories that is whether you have the disease or you do not have the disease. This is computed by a function we have defined in the flask app which passes the user input as a numpy array to the ML Modelandreturnstheoutputastheprediction.
Inthefigureaboveyoucanseethatweareloadingoursaved picklemodelsandrenderingthembasedonthelengthofthe user input. This is a very logical and effective way of managingthemodelsandmakingsurethecorrectmodelis loadedaccordingtotheuserinput.
6. RESULTS
Figure 14 AccuracyScoreofLiverModel
D.2 Accuracy Score
TheAccuracyscorefortheKidneymodelisinchperfect.The model after being implemented in the Random Forest Classifiergivesa100%Accuracy.
Figure 18 App.pyfile
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page730
IntheFigureattachedabove,thereisaformwhichaccepts theuserinputandpassesittotheflaskfileviz.app.py.After theuserhasgiveninputtothewebsitetheinputisprocessed andtheresult/outcomeisdisplayed.
7.
Fromthiswecanconcludethatallthe4machinelearning models are working properly while being integrated in a website. The users can give their input and successfully obtain the result/prediction after pressing the submit button.ThemaintakeawaypointfromthisProjectistheway inwhichtechnologieslikeMachineLearningcanpredictand have influence in our day to day lives. A few years Ago, it would’ve been difficult to predict health of people consideringthereissomuchatstake.Butthankstoamazing researchandtechnologythatwehaveatourdisposaltoday it is done very easily. It is always a pleasure to learn new skillsandimplementthemonaprojectthataddssomevalue tobettermentandthewell beingofthesociety.
Figure 22 OutputShowingthattheuserisHealthy Theabove2FiguresthatisFigure21and22aretheoutput windows.Thisiswhattheuserseesasanoutputafterthe inputisprocessed.Theoutputisintheformofamessage justtokeepitassimpleaspossible. CONCLUSION
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page731
[4] G.Chenetal.,"PredictionofChronicKidneyDisease Using Adaptive Hybridized Deep Convolutional Neural NetworkontheInternetofMedicalThingsPlatform,"inIEEE Access, vol. 8, pp. 100497 100508, 2020, doi: [5]10.1109/ACCESS.2020.2995310.P.Chittoraetal.,"Prediction of Chronic Kidney Disease AMachineLearningPerspective,"inIEEEAccess, vol. 9, pp. 17312 17334, 2021, doi: 10.1109/ACCESS.2021.3053763.
ACKNOWLEDGEMENT
Figure 20 DiabetesPredictionForm
Figure 19 LandingPageoftheWebsite
REFERENCES
Figure 21 OutputShowingtheuserhastheDisease.
[1] A. Gavhane, G. Kokkula, I. Pandya and K. Devadkar, "PredictionofHeartDiseaseUsingMachineLearning,"2018 Second International Conference on Electronics, CommunicationandAerospaceTechnology(ICECA),2018, pp.1275 1278,doi:10.1109/ICECA.2018.8474922.
[3] A.H.Chen,S.Y.Huang,P.S.Hong,C.H.ChengandE. J. Lin, "HDPS: Heart disease prediction system," 2011 ComputinginCardiology,2011,pp.557 560.
[2] A.SinghandR.Kumar,"Heart Disease Prediction Using Machine Learning Algorithms," 2020 International ConferenceonElectricalandElectronicsEngineering(ICE3), 2020,pp.452 457,doi:10.1109/ICE348803.2020.9122958.
WewouldliketothankourCollegeVishwakarmaInstituteof Technology, Pune for providing us with a platform to conductourresearchandimplementourproject.Wewould liketoalsoexpressgratitudetoourProjectguideProf.Dr. Vaishali Jabade ma’am for her constant guidance and supporttous.
In the above Figure you can see the landing page of our website. Here we have displayed various resources and information regarding diseases that we predict. We have includedseveralbloglinksandembeddedvideostoeducate theusersaboutDiabetes,Kidney,LiverandHeartDiseases.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page732 [6] A. Anand and D. Shakti, "Prediction of diabetes based on personal lifestyle indicators," 2015 1st International Conference on Next Generation Computing Technologies (NGCT), 2015, pp. 673 676, doi: [7]10.1109/NGCT.2015.7375206.A.MirandS.N.Dhage,"DiabetesDiseasePrediction Using Machine Learning on Big Data of Healthcare," 2018 Fourth International Conference on Computing CommunicationControlandAutomation(ICCUBEA),2018, pp.1 6,doi:10.1109/ICCUBEA.2018.8697439. [8] S.Sontakke,J.LohokareandR.Dani,"Diagnosisof liverdiseasesusingmachinelearning,"2017International ConferenceonEmergingTrends&InnovationinICT(ICEI), 2017,pp.129 133,doi:10.1109/ETIICT.2017.7977023.