Neural Net: Machine Learning Web Application

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

Neural Net: Machine Learning Web Application

Arpan Dhar, Nandita Ayyar, D. Varun, G. Annapoorani Student, Department of ECE Student, Department of ECE Student, Department of ECE Assistant Professor, Department of ECE SRM IST, Ramapuram, Chennai, India ***

Abstract The Neural Net is a “machine learning” embedded web application. Its sole purpose is to implement the applications of machine learning and making it easily available for users in just one click. In this, Machine learning algorithms like linear regression and logistic regression have been used. There are many live web applications like “covid predictor”; “used car price predictor” on our web app. Many more projects are coming soon like “stock trend predictor” etc. Model has been trained to give minimal error. (Please note: These models cannot be 100% accurate)

Key Words: machine learning, deep learning, Linear regression, logistic regression, front end, back end, web application, prediction, Covid19 Prediction, used car price prediction

1. INTRODUCTION

NeuralNetitisawebbasedapplication.Themainpurposeofthiswebapplicationistointegratethefacilitiesofwebsitesand web apps through machine learning. A web app is similar to website but a web app serves a specific purpose. Different machinelearningalgorithmsarebeingusedinthisapplicationwhichcaneasethedaytoday lifeofitsusers.Theinitialphase of the web application doesn’t have many features but with time and more expertise we are aiming to make this anall in one machine learning web app.

With the advancements in this era, technologies such as machine learning and deep learning are used in day to day life applications without even realization. Here in this prototypeof this web app we’ve provided the facility of used car price predictor and covid predictorUsed car predictor takesdifferentinputs from the user and runs an algorithm to calculate the price in which the car can be sold. Covid predictor also works in a similar manner. It also takes inputs from the user in the formofsymptomsandpredicttheprobabilityofhim/herof beinginfected.Although,afterpredictiontheuserisadvisedtodo further investigations under medical supervision since this web app just merely works on predictions based on external symptoms no on the basis of any test.

Machine Learning:

Machinelearning (ML) is the study of computer algorithms that may better themselves over time by gaining experience and using different data. Machine learning algorithms create a model based on training data to make predictions or judgments without having to be explicitly programmed to do so.

Types of Machine Learning Techniques

1.1.Supervised Learning:Itisatypeofmachinelearning whichincludesfeatures andlabelsfortrainingthemodel.Features aretheinputsonwhich themodel istrainedandlabelsare the outputs which the model predicts. Here, the whole dataset is dividedinto trainsetand testsetand sometimesinto train set, validation set and test set. Linear regression isanexample ofsupervised learning

1.2.Unsupervised Learning:

Here,no labels are provided to the model. Based ontheinputs/featuresprovidedtothemodel,itappliessomealgorithm onit.Forexample,let’sassumethatwereinapartyandtherearepeoplechatteringandsomemusicisalsobeingplayedinthe background. Now since there is no clear labels thus the algorithm will try to separate the incoming inputs (sounds) on the

© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal

Page4015

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056 Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

basis of similarities. Thus it will divide the whole noise into 2 broad categories first, the sound of the music and second the voice of people chattering.

1.3.Reinforcement Learning:

The best possible way to get the desired output, is Reinforcement Learning. It has different probabilities of output for the same problem and it keeps on learning until it gets the desired output. Even if there is no training dataset like supervised learning, it can predict the output by trial and error basis. It is similar to sequential circuits, where your output depends on previous stage output and current input.

Fig. 1. Varioustypesof“MachineLearningTechniques”

2. MACHINE LEARNING TASKS ANDALGORITHMS USED IN PROJECT

2.1. Logistic regression (LR) :

Toestimateprobabilities,logisticregressioncommonlyemploysalogisticfunction,oftenknownasthemathematicallydefined sigmoid function. It can be used to solve both classification and regression problems, but the former is the more popular application.

This algorithm has been used in the covid predictor to predict the probability of the user being infected. Later on,the probability has been converted into percentage to make theoutput more comprehendible. Since it is not necessary that the bestfittedlineisalwaysastraightline,itmayhappenthatthe outputsarebestfittedbysomeirregularshapelikeacircleor somelogfunction,inthesecaseslogisticregressionismostlyused.

2.2. Linear Regression

Thedependentvariableiscontinuousinthisapproach,the independentvariable(s) mightbecontinuousor discrete,and the regressionlineislinear.Linearregressionusesthebestfitstraightlinetobuildalinkbetweenthedependentvariable(Y)and oneormoreindependentvariables(X)(alsoknownastheregressionline).Theequationsthatdefineitareasfollows:

where a is the intercept, b is the slope of the line, and e istheerrorterm.

© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal

Page4016

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

Fig. 2. SupervisedLearning

Fig. 3. UnsupervisedLearning

Fig. 4. Reinforcement Learning

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal

Page4017

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

Fig. 5 LogisticRegression

3. SIMULATION AND TECHNOLOGIES USED

Frontendlanguage:html,CSS,JavascriptBackendlanguage:node.js,Javascript,flask

Webframework: flask (python)

ML library used: pandas, numpy, tensorflow, scikit learnDataset for Covid Predictor randomly generated since there is no suchofficialavailable

DatasetforCarPricePredictor:Quickr

4. Front end:

The front end is the user interface part of the web app which the user interacts with. The designes and looks of the webapp comes under the front end. There are many languages to do front end designing but we’ve majorly used HTML , CSS, JavaScript,Node.js,Express.js.

HTML Itismostlyusedincreatingtheskeletonofthewebapp.

CSS Itismostlyusedindesigningthewebappforattractivecolorsandfonts.

JavaScript Thisisresponsibleforthefunctioningofthewebpage.Dropdownmenusandotherinteractivemovementstakes placeunderscripttagofJavaScript.

5. Back end:

Amachinelearningalgorithmistrained using a containingsymptoms of peopleand whether they are testedpositive or not. Sincerightnownogovtapprovedisavailableforpeople,wehavecreatedarandomhavingaround2500inputsandoutputsto trainthemodelThismachinelearningalgorithmuseslinearregressiontotrainthemodelwithminimalerror.

Fortheotherproject,i.e.usedcarpricepredictor,themachinelearningalgorithmistrainedusingadatasetobtainedfroman open source website that contains many input values that aids in the prediction of the car price based on various factors on whichitdepends.Here,logisticregressionisusedtotrainthemodel

Itisresponsibleforall the background functioning ofa website whichtheusercannotsee. Suppose weclick ona link inthe web page and we are redirected to that particular page, this process comes under backend of the website. There are many languages used for backend. We’ve used Express.js for writing backend codes. Further the machine learning algorithms are

© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page4018

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

integratedwiththewebpageusing Python supportedFlaskwhichworksasabridgebetweenthewebappandthemachine learningmodels.

6. Result and Discussion:

In the application Neural Net, HTML is used for the basic structuring and CSS for styling. JavaScript is used as the scripting language.

Onceauseropensthedomainwww.neuralnet.co.inthehomepageofNeuralNetopensup.

TheHomePagedisplaysfourmoreoptions.Byclickingontheseoptions,theuserwillberedirectedtotherespectivepages.

Thisfrontendwebsiteiscompletelyresponsive.Itcanbeviewedinsplitscreenonintablets,desktopsaswellasmobiles.On mobileandtabletstheuserwillgetadropdownmenufortheseoptions

Fig. 6 InterfaceofWebApp

Theoptionsareasfollows:

Home:Displayshomepageofthewebapplication.

About:Givesdetailedinformationaboutthecreators. Project:Containsalltheservicesprovidedbythewebapplication.

ContactUs:Providestheuserwithafeedbackformtocontactthedomainadministrators.

Asofnow,thedomainonlydisplaysaprototypeofthewebapplication,henceitcontainsonlyfewservices.

2022,

Fig. 7 Interfaceoftheservicepageofwebapplication

Factor value: 7.529

9001:2008

4019

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

Working :

CovidPrediction

Once the user clicks on the Covid Prediction service, a form is displayed on the screen with multiple questions for them to answer.Thesequestionsarefedtothemachinelearningalgorithmasfeatures/inputs,andthelabel/outputisdisplayed.

Features/input:symptoms

Labels/output:probabilityofthembeinginfected

FormComponents:Features/input/symptoms:

1. Do you have fever? The user has to enter their body temperature. The input will be fed to machine learning model in continuousvalues.

2.Enteryourage:TheUserhavetoentertheirage.Theinputwillbefedtomachinelearningmodelascontinuousvaluefrom1 to 100.

3. Do you have body pain? The User will be provided with 2 options, yes or no. The input will be fed to machine learning model as discrete 1 for yes and 0 for no.

4.Doyouhavea runnynose? TheUserhavetoenter whethertheyhaverunnynoseornot.Theinputwillbefedto machine learning model as discrete 1 for yes and 0 for no.

5. Do you have breathing difficulties? The User will be provided with 3 options to enter, whether they have breathing difficulties or not. The input will be fed to machine learningmodelasdiscrete1forextremedifficultyand0formilddifficulty and 1fornodifficulty.

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

Labels/output:0% 100%(ofgettingcovid19)

ThisprojectwillhelpuserstopredictchancesofapersonbeinginfectedbythenovelCovid19virus.Thiscanhelpinspeeding upthetestingprocessandisolationasitwill beeasiertodiagnosetheailment.Peoplehavinghighpercentage outputshould betreatedearliercomparedtothosehavinglowpercentage output.

A. Used Car Price Predictor:

Itisanotherserviceofneuralnetwebappwhichuseslinearregressiontopredictthepriceofanyused car. Inputstakenby thisservicearecarbrand,car model,yearof purchase,howmanykilometersdriven,petrol/dieselvariantetc.After entering theinputs, it predicts the output price of the caratwhichitshouldbesold.Thismodelhasbeentestedforminimalerror and maximum accuracy. This model has been trained on a dataset containing more than 890 data of differentmodels of differentcar brands.

Pythonlibrariesused:Pandas,numpy,matplotlib,scikit learn

Linearregressionclassifierisusedtotrainthemodel.Theoutputcannotbepredictedwith100%accuracyFeatures/Input :

1.Selectthecompany:Awiderangeofcarbrandswillbeavailableforoptions

2.Selectthemodel:Afterselectingtheparticularcarbrandtheuserwillbeprovideswithawiderangeofcarmodelsunderthat brand

3.Selecttheyearofpurchase:Userhastotheyearofpurchaseofthecar

4.Selectthefueltype:Twooptionsareprovidedpetrolordiesel.

5. Enter the number of kilometers that the car has been driven: User has to enter the total distance the car has already travelled.

Flaskhasbeenusedtodeploythemachinelearningmodelintothewebapplication

© 2022, IRJET

Fig 10. Interfaceoftheusedcarpricepredictor

Fig. 11.Outputoftheusedcarpricepredictor

Impact Factor value: 7.529

9001:2008

4021

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN:2395-0072

8. CONSLUSIONS

The website displayed the web app with projects in workingcondition with absolute proficiency and minimum errors. The Covid19predictorandUsed Carpricepredictorworkfinewhentheuserinputsthedataandclicksoncalculate.The resultis thendisplayedasanoutputonthescreen.Thewebapplicationusesmachinelearningalgorithmstosuccessfullygiveadesired outputtotheuser.

REFERENCES

[1] NilssonN.“Introductiontomachinelearning[webdrafted.]”, 1996

[2] MiguelGrinberg“FlaskWebDevelopment:DevelopingWebApplicationswithPython”, 2014

[3] Iqbal H. Sarker “Machine Learning: Algorithms, Real WorldApplicationsandResearchDirections”2021

[4] MaximeJan,NastassiaGobet,ShanazDiessler, Paul Franken, Ioannis Xenarios “Amulti omics digital research object forthegeneticsofsleepregulation”,ScientificData,2019

[5] J.R.Castillo Saenz,N.Nedev,B.Valdez Salas,M.Bernecheaetal.“Effectofoxidationtemperatureonthepropertiesof NiOxLayersforapplicationinopticalsensors”,ThinSolidFilms,2021

BIOGRAPHIES

Arpan Dhar

Founder ofNeuralNetWebApp HaveexpertiseinMachineLearning, Data Science &Full StackDevelopment.

2022, IRJET

D. Varun

Co-Founder ofNeuralNetWebApp HaveexpertiseinFront endDevelopment.

Nandita Ayyar

Co Founder ofNeuralNetWebApp ContentCreatorandNon TechnicalRoles

Impact Factor value: 7.529 | ISO 9001:2008