House Price Prediction Using Machine Learning Via Data Analysis

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

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072

House Price Prediction Using Machine Learning Via Data Analysis

Abstract - Research teams are increasingly adopting machine learning modelstoexecuterelevantproceduresinthe field of house price prediction. As some research did not take into account all available facts, influencing house price forecast and produces inaccurate results. The House Price Index (HPI) is a popular tool for estimating changes in house costs depending on factors such as location, population, industrial growth, and economic prospects. This paper givesa general overview of how to anticipate priceofhousesbasedon customer requirements utilizing traditional data and advanced machine learning models, together with regression techniques and python libraries. The effectiveness of our analysis is confirmed by the usage of ANN (Artificial Neural Network), locational attributes, structural attributes, and data-mining’s capacity to extract knowledge from unstructured data. This housing price forecast model for Tier1 cities, with an accuracy of more than 85%, offers enormous benefits, particularly to buyers,developers,andresearchers,as prices continue to fluctuate.

Key Words: House PricePrediction,RegressionModels, Data Analysis, Machine Learning using Python, Algorithms.

1. INTRODUCTION

Asweallknow,ahouseisabasichumanneed,andcostsfor themvaryfromplacetoplacedependingonamenitieslike parkingandneighbourhood.The[2]housingmarketshavea favourable effect on a nation's currency, which is a significantfactorinthenationaleconomy.Everyyear,there isagrowthinthedemandforhomes,whichindirectlydrives up home prices. The issue arises when there are many factors, such as location and property demand, that could affect the house price. As a result, the majority of stakeholders,includingbuyers,developers,homebuilders, andtherealestateindustry,wouldliketoknowtheprecise characteristicsortheaccuratefactorsinfluencingthehouse pricetoassistinvestorsintheirdecisionsandtoassisthome buildersinsettingthehouseprice.Viewing[5]transaction records may help buyers determine whether they were givenafairpriceforahomeandsellersdeterminetheprice at which a home can be sold in a certain area. Finding a workablepredictionmethodisthereforeincreasinglycrucial becausetheemploymentofasingleclassifierislimited.

1.1 Motivation

Many people, whether wealthy or middle class, are concerned about house prices. It is possible to establish a mechanism to predict correct prices based on previous buy/sellvaluesbecauseonecanneverappraiseorestimate the pricing of a house based on the neighborhood or amenities offered. Making sure everyone can purchase a homeatthebestpriceisthekeygoal.

1.2 Objectives

Thisprojectisbeingputforth[1]toforecasthouseprices and to acquire better and accurate results. It will use a varietyofregressionmethodstoprovidethemostprecise and accurate findings. Python programming language is employed for machine learning in order to complete this operation.Todeterminewhichregressionmethodproduces themostpreciseandaccurateresults,thestackingalgorithm isperformedtomultipleregressionalgorithms.

1.3 Problem Statement

Thereisnoproperpricefixationbecausepricesintier-1 cities are always changing and dependent on a variety of factors,includinglocation,population,andpotentialfuture projects. Purchasers, developers, and researchers are all impactedbythis.Usersneedasuitableplatformtoobtain accurate prices based on historical selling price patterns, employingavarietyoffeaturesandgatheringinformation fromnumeroustier-1citylocations.

1.4 Machine Learning using Python

Python is a sophisticated, widely used programming language. In 1991, "GUIDO VAN ROSSUM" invented it. Numerous libraries, including pandas, numpy, SciPy, matplotlib,etc.,aresupportedbyPython.ItsupportsXlsx, Writer,andX1Rd,amongotherpackages.Complexscienceis performed extremely effectively using it. There are numerousfunctionalPythonframeworks.

Machinelearningisabranchofartificialintelligencethat allows computer frameworks to pick up new skills and enhance their performance with the help of data. It is employedtoresearchthedevelopmentofcomputer-based algorithmsformakingpredictionsaboutdata.Providingdata isthefirststepinthemachinelearningprocess,afterwhich thecomputersaretrainedbyusingavarietyofalgorithmsto

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

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072

create machine learning models. Software engineering's branch of machine learning has significantly altered how peopleanalysedata.

2. RELATED WORKS

Accordingtoresearch[4],theHousePriceIndex(HPI)isused inmanynationstogaugevariationsinthecostofresidential realestate,toforecasttheaveragepricepersquarefootof each house, the dataset including a large number of data pointsandawiderangeoffactorsindicatingthehousevalues tradedinpreviousyearswasemployed.Totidyupthedata, specifyminimalvaluesfortheparameters"area"and"price." Prior to creating a regression model, exploratory data analysisiscrucial.Researchersareabletoidentifyimplicit patternsinthedatainthisway,whichhelpsthemselectthe bestmachinelearningtechniques.

Basedon[6]MultipleLinearRegression,themethodology [MLR] The most prevalent type of linear regression is multiplelinearregression.WiththeuseofEquations1and2, the[MLR]multiplelinearregressionisemployedasaforecast predictor to demonstrate the relationship between a continuousdependentvariableandoneormoreindependent variables:E(Y|X)=1+1X1++pXp -(1)Yj=1+1Xj,1++ pXj,p + j - (2). The various regression techniques might include [LR] One form of linear regression that makes advantage ofshrinking is the lasso regression. [ER]In this method, the regularization issue is resolved via elastic net regression.Fora non-negative, preciselybetween 0and1, anda1.TheGradientBoostingalgorithm(GBR)isamachine learning method for resolving issues with regression and classification.Asaresult,apredictionmodeliscreatedthat combinesalloftheweakpredictionmodels,mostlydecision trees.AdaBoostingRegression(AR)isaregressionapproach designedtocombine"simple"and"weak"classifierstocreate a"strong"classifier.

The[10]CNN-basedpredictionmodelforhomepricesisa typeoffeed-forwardneuralnetworkwithadeepstructure and convolution processing, which is one of the representativedeeplearningalgorithms.Intheexperimental phase,theCNNmodelisconstructedusingtheTensorFlow framework. The activation function for the model's two convolutional layers is the Relu function, and the dropout methodisalsousedtopreventover-fitting.Resultsfromthis CNNmodelexperimentis98.68%accurate.Although[11]the accuracy, precision,specificity,andsensitivityare the four criteria used to assess the success of machine learning systems. The two discrete values 0 and 1 are regarded as distinctclassesinthework.Iftheclassvalueis0,weassume thatthehouse'spricehasreduced,andiftheclassvalueis1, weassumethatthehouse'spricehasgrown.Theaccuracy levels attained with this technique for various machine learning methods are as follows: Random Forest - 86%, SupportVectorMachine-82%,andArtificialNeuralNetwork -80%.

Asin[12]oneofthevaluesofthetaxobject(NJOP)ofthe landandthevalueofthetaxobjectofthebuildingarethe factorsthatcanbequantitativelycalculatedtodeterminethe selling price of the home. Numerous elements, like the building's age and strategic placement, have an impact on both criteria. To acquire the best prediction accuracy, the initialhousepricepredictionisdifficultandcallsforthefinest methodology.Fuzzylogicbecomesoneofthestrategiesthat canbeemployedinsolvingtheproblemofestimatingthesale price of a house that has an uncertainty parameter. For predictingresidentialpropertyprices,predictionscanalso makeuseoftheK-NearestNeighborsmethodinadditionto fuzzylogicandartificialneuralnetworks.

In [13] the nonlinear model for real estate's price variation prediction of any city, based on leading and concurrent economic indices, is established using two techniques,similartothoseusedinbackpropagationneural network (BPN) and radial basis function neural network (RBF).Theoutcomesofthepredictionsarecontrastedwith thePublicHousePriceIndex.Thetwoindicesoftheprice fluctuationthatarechosenastheperformanceindexarethe meanabsoluteerrorandrootmeansquareerror.

Similar to [14], the GA-BP model, which combines the geneticalgorithm(GA)andbackpropagationneuralnetwork (BPNN), was used to predict the housing price along the urbanrailtransitline.Thisstudyexaminedtherelationship betweenaccessibilityalongtheurbanrailtransitlineandthe change in housing price. By using the cost of residential areasnearabusymetrolineasanexampleandcontrasting theperformanceoftheGA-BPmodelandtheBPmodel,the model'sdependabilitywasconfirmed.Itwasdiscoveredthat themeansquareerrorofthepredictionoutcomesismuch lowerandthattheaverageabsoluteerrorrateoftheGA-BP model is6.91%,whichis6.45%lowerthanthatofthe BP model. As can be shown, the GA-BP model-based price prediction algorithm for those residential areas near the urbanlineisaccurateandcompletelyexploitsthepotential relationship between the affordability of housing and the accessibilityoftheurbantransportationnetwork.

As with reference to [15], an unsupervised learnable neuronmodel(DNM)byincludingthenonlinearinteractions betweenexcitationandinhibitionondendritesisused.We fittheHousePriceIndex(HPI)datausingDNM,afterwhich we project the trends in the Chinese housing market. We compare the performance of the DNM to a conventional statisticalmodel,theexponentialsmoothing(ES)model,in ordertoconfirmtheefficacyoftheDNM.Theaccuracyofthe twomodels'forecastsisassessedusingthreequantitative statisticalmetrics:normalizedmeansquareerror,absolute percentageoferror,andcorrelationcoefficient.Accordingto experimentalfindings,thesuggestedDNMoutperformsthe ESineachofthethreequantitativestatisticalcriteria.

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page974

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

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072

3. IMPLEMENTATION AND WORKING

3.1 Architectural Diagram

The[7]thebasicstructuralworkingmethodologyisbased ontheaboveflowchart.

3.2 Data Collection and Data Cleaning

There [3] are several methods and procedures used in dataprocessing.DatafromrealestatepropertiesinTier-1 citieswasgathered fromseveral real estate websites. The information would include attributes like location, carpet size,built-uparea, property age,zipcode, etc. Weneed to gather structured and organized quantitative data. Before beginning any machine learning research, data must be collected.Withoutdatasetvalidity,therewouldbenosense inevaluatingthedata.

Errors are found and removed throughout the data cleaningprocessinordertomaximizethevalueofthedata. To guarantee that accurate and correct information is available from the record set, table, or dataset, it replaces untidy information. The main goal of data cleaning is to provideadynamicestimationofinformation

determiningthebestwaystohandlemissingdata.Sinceitis challenging to impute these missing values with an acceptablelevelofaccuracy,columnswithmorethan55%of theirvaluesmissingareeliminatedfromtheoriginaldataset. Additionally,theresultvariable'svaluesaremissingfroma largenumberofrows(Price).Theobservationswithmissing values in the Price column are eliminated since the imputation of these values can enhance bias in the input data.

3.4 Data Visualization

Duetoitscapacitytosuccessfullyinvestigatechallenging conceptsandfindnovelpatterns,datavisualizationisused inmanyhigh-qualityvisualrepresentations.

3.5 Cross Validation and Training the Model

Inordertotrainamachinelearningmodelusingasubset of the dataset, cross validation is performed. Training is importanttoobtainaccuracywhendividingdatasetsinto "N" sets for assessments of the model that has been constructed.

Wemustfirsttrainthemodelbecausethedataisdivided into two modules: a test-set and a training-set. The target variableisapartofthetrainingset.Thetrainingdatasetis subjected to the decision tree regressor method. Using a tree-like structure, the decision tree creates a regression model.

3.6 Testing and Integration with UI

AwebframeworkFlask,givesyouthetechnology,tools, andlibrariesnecessarytocreateawebapplication.Flaskisa framework primarily used for integrating Python models becauseitissimpletobuildtogetherroutes.

Housepricesareforecastedusingthetrainingmodeland a test dataset. The front end is then connected with the trained model using Python's Flask framework. After creatingthemodelandsuccessfullyproducingthedesired result,theintegrationwiththeuserinterface(UI)isthenext stage,andflaskisemployedforthis.

3.3 Data Preprocessing and Analysis of Data

The[8]dataset needs to be pre-processed before using models to predict house prices. First, a missing data investigation is carried out. A number of missing patterns are systematically evaluated since they are crucial in

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

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072

homes'salesprices.Thesalespriceshavebeendetermined more precisely and accurately. The public would benefit greatlyfromthis.Variousdataminingmethodsareusedin Python to produce these findings. It is important to think aboutandaddressthenumerousaspectsthathaveanimpact onhomepricing.Wereceivedhelpfrommachinelearningto finishourassignment.Datacollectingisdonefirst.

Then, data cleaning is done to make the data clean and remove all of the errors. The data pre-processing is then completed. Then, several graphs are made using data visualization. This has shown how data is distributed in several ways. Furthermore, the model is prepared for use andtested.Somecategorizationtechniqueswerediscovered tohavebeenusedonourdataset,whereasothershadnot. Therefore,thealgorithmsthatwerenotbeingusedonour housepricingdatasetwereremoved,andeffortsweremade toincreasetheaccuracyandprecisionofthealgorithmsthat were.Auniquestackingapproachissuggestedinorderto increasetheprecisionofourclassificationsystems.Inorder to get better outcomes, it is crucial to increase the algorithms'accuracyandprecision.Thepeoplewouldnotbe abletoestimatethesalespricesofhousesiftheresultswere inaccurate. Data visualization was also used to improve accuracy and outcomes. Various algorithms are used to determine the homes' sales prices. The sales prices have beendeterminedmorepreciselyandaccurately.Thepublic wouldbenefitgreatlyfromthis.

4. RESULTS

Python-baseddataminingtechniquesareusedtoproduce thedesiredresults.Numerousvariablesthathaveanimpact on home pricing are taken into consideration and further developed.Theideaofusingmachinelearningtocarryout the desired task has been considered. Data collection is startedfirst.Then,datacleaningiscarriedouttomakethe data clean and error-free. The next step is data preprocessing.Thedistributionofdatainvariousformsisthen intendedtobedepictedthroughthecreationofvariousplots usingdatavisualization.Intheend,thebusinesscostsofthe homeswerecalculatedprecisely Inadditiontoemploying regressiontechniques,severalclassificationalgorithmsare takenintoaccountandusedonourhousepricingdataset, including the SVM algorithm, decision tree algorithm, RandomForestclassifier,etc.awaythatwouldassistpeople inpurchasinghomesatapricethatisaffordableandwithin theirmeans.Variousalgorithmsareusedtodeterminethe

5. FUTURE SCOPE

ThefutureobjectiveistoexpandthedatasettootherIndian statesandcities,asitcurrentlyonlycomprisesTier-1cities. We'll be integrating map into the system to make it even more informative and user-friendly. As [20] numerous importantfactorsinfluencepropertyvalues.Itisagoodidea to add extra elements if statistics are available, such as income, salary, population, local amenities, cost of living,

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

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072

annual property tax, school, crime, and marketing information.Inthenearfuture,we'llgiveacomparisonof the price projected by the system and the price from real estatewebsiteslikeHousing.com,magicbricks.cometc.,for the same user input. We will also suggest real estate properties to the consumer depending on the anticipated pricingtofurthersimplifythingsforthem.

6. CONCLUSION

Thiswork[16]usesvariousmachinelearningalgorithmsto undertakeananalyticalanalysisoftheeffectsofrealestate market fluctuations on real estate health and trends. The abilitytoestimatehousepricesiscrucialforminimisingthe effects of property valuation and economic expansion in complicatedrealestatesystems.Amongtheseveralmachine learningmethods,XGBoostisappliedtoenhancehomeprice predictioninintricaterealestatesystems.The[17]trialand errorisusedtoidentifythebestANNmodel.But[18]ANN hasitsdrawbacks.Theinconsistentnatureofthefindingsis onemajorissue.TheANNmodeliscapableofself-learning during the training phase and modifying the weights appropriatelytoreducetheerror.Asaresult,itisimpossible totellwhetherthemodelproducedthebestresult.Aseries oftestswererunonapublicrealestatedatasetinorderto assessandcontrastthesuggestedmodel.Accordingtothe experimental data, the XGBoost model is more successful thanotherbaselinemachinelearningpredictionmethodsat predictinghomepricesforrealestate,attaining89%ofthe measure.Bothmarketplayersandbankscanusethisresult. As[19]themarketplayersareconstantlycuriousaboutthe likelihoodoftheirhomesselling.However,ifaborroweris unable to repay the loan, it would be fascinating for the banks to know how much chance they have of selling the homes they have taken as collateral. The maintenance of financialstabilitydependsheavilyonthelatter.

7. ACKNOWLEDGEMENT

We thank, Dr. Thomas P John (Chairman), Dr. Suresh VenugopalP(Principal), DrSrinivasaHP(Vice-principal), Ms.SumaR(HOD–CSEDepartment),Dr.JohnTMesiaDhas (Associate Professor & Project Coordinator), Ms. Bhavya Balakrishnan (Assistant Professor & Project Guide), Teaching & Non-Teaching Staffs of T. John Institute of Technology,Bengaluru–560083.

REFERENCES

[1] MansiJain,HimaniRajput,NehaGarg,PronikaChawla, “Prediction of House Pricing Using Machine Learning with Python,” IEEE Xplore Part Number: CFP20V66ART; ISBN: 978-1-7281-4108-4, DOI: 10.1109/ICESC48915.2020.9155839

[2] Nor Hamizah Zulkifley, Shuzlina Abdul Rahman, Nor Hasbiah Ubaidullah, Ismail Ibrahim, “House Price PredictionusingaMachineLearningModel:ASurveyof

Literature,” I.J. Modern Education and Computer Science,2020,6,46-54,DOI:10.5815/ijmecs.2020.06.04

[3] AlishaKuvalekar,ShivaniManchewar,SidhikaMahadik, Shila Jawale (Guide), “House Price Forecasting Using MachineLearning,”Proceedingsofthe3rdInternational Conference on Advances in Science & Technology (ICAST)2020at:https://ssrn.com/abstract=3565512

[4] QuangTruong,MinhNguyen,HyDang,BoMei,“Housing Price Prediction via Improved Machine Learning Techniques,” 2019 International Conference on Identification, Information and Knowledge in the Internet of Things (IIKI2019) - Procedia Computer Science 174 (2020) 433–442, DOI:10.5815/ijmecs.2020.06.04

[5] Pei-YingWang,Chiao-TingChen,Jain-WunSu,Ting-Yun Wang,Szu-HaoHuang,“DeepLearningModelforHouse Price Prediction Using Heterogeneous Data Analysis AlongwithJointSelf-AttentionMechanism,”IEEEAccess –DOI:10.1109/ACCESS.2021.3071306

[6] CH.RagaMadhuri,AnuradhaG,M.VaniPujitha,“House Price Prediction Using Regression Techniques: A ComparativeStudy,”IEEE6thInternationalConference on smart structures and systems ICSSS 2019, DOI: 10.11098/ICSSS.2019.8882834

[7] P. Durganjali, M. Vani Pujitha, “House Resale Price Prediction Using Classification Algorithms,” IEEE 6th International Conference on smart structures and systems ICSSS 2019 - 978-1-7281-0027-2/19/$31.00 ©2019IEEE,DOI:10.1109/ICSSS.2019.8882842

[8] The Danh Phan, “Housing Price Prediction using MachineLearningAlgorithms:TheCaseofMelbourne City, Australia,” 2018 International Conference on Machine Learning and Data Engineering (iCMLDE)978-1-7281-0404-1/19/$31.00 ©2019 IEEE, DOI 10.1109/iCMLDE.2018.00017

[9] Yajuan Tang, Shuang Qiu, Pengcheng Gui, “Predicting HousingPriceBasedonEnsembleLearningAlgorithm,” DOI:10.1109/IDAP.2018.8620781

[10] Yong Pia, Ansheng Chen, Zhendong Shang, “Housing Price Prediction Based on CNN,” 9th International Conference on Information Science and Technology (ICIST), Hulunbuir, Inner Mongolia, China, DOI: 10.1109/ICIST.2019.8836731

[11] Debanjan Banerjee, Suchibrota Dutta, “Predicting the Housing Price Direction using Machine Learning Techniques,”IEEEInternationalConferenceonPower, Control, Signals and Instrumentation Engineering (ICPCSI-2017)-978-1-5386-0814-2/17/$31.00©2017 IEEE,https://doi.org/10.1109/ICPCSI.2017.8392275

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page977

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

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072

[12] Muhammad Fahmi Mukhlishin, Ragil Saputra, Adi Wibowo,“PredictingHouseSalePriceUsingFuzzyLogic, Artificial Neural Network and K-Nearest Neighbour,” 20171stInternationalConferenceonInformaticsand Computational Sciences (ICICoS) - 978-1-5386-09033/17/$31.00 © 2017 IEEE, DOI: https://doi.org/10.1109/ICICOS.2017.8276357

[13] LiLiandKai-HsuanChu,“PredictionofRealEstatePrice VariationBasedonEconomicParameters,”Proceedings ofthe2017IEEEInternationalConferenceonApplied SystemInnovationIEEE-ICASI2017-Meen,Prior&Lam (Eds) - ISBN 978-1-5090-4897-7, DOI: https://doi.org/10.1109/ICASI.2017.7988353

[14] LiRuo-qiandHuJun-hong,“PredictionofHousingPrice Along the Urban Rail Transit Line Based On GA-BP Model andAccessibility,”2020IEEE5th International ConferenceonIntelligentTransportationEngineering978-1-7281-9409-7/20/$31.00 ©2020 IEEE, DOI: https://doi.org/10.1109/ICITE50838.2020.9231460

[15] Ying Yu, Shuangbao Song, Tianle Zhou, Hanaki Yachi, ShangceGao,“Forecasting HousePriceIndexofChina Using Dendritic Neuron Model,” 978-1-5090-34840/16/$31.00 ©2016 IEEE, DOI: https://doi.org/10.1109/PIC.2016.7949463

[16] Bandar Almaslukh, “A Gradient Boosting Method for EffectivePredictionofHousingPricesinComplexReal Estate Systems,” 2020 International Conference on TechnologiesandApplicationsofArtificialIntelligence (TAAI),DOI:10.1109/TAAI51410.2020.00047

[17] WanTengLim,LipoWang,YaoliWang,andQingChang, “HousingPricePredictionUsingNeuralNetworks,”2016 12thInternationalConferenceonNaturalComputation, FuzzySystemsandKnowledgeDiscovery(ICNC-FSKD)978-1-5090-4093-3/16/$31.00 ©2016 IEEE, DOI: 10.1109/FSKD.2016.7603227

[18] Lipo Wang, Fung Foong Chan, Yaoli Wang, and Qing Chang,“PredictingPublicHousingPricesUsingDelayed Neural Networks,” 2016 IEEE Region 10 Conference (TENCON) Proceedings of the International Conference - 978-1-5090-2597-8/16/$31.00 ©2016 IEEE,DOI:10.1109/TENCON.2016.7848726

[19] CeyhunAbbasov,“Thepredictionofthechanceofselling ofhousesasthefactoroffinancialstability,”2016IEEE 10th International Conference on Application of InformationandCommunicationTechnologies(AICT), DOI:10.1109/icaict.2016.7991786

[20] SifeiLu,ZengxiangLi,ZhengQin,XuleiYang,RickSiow MongGoh,“AHybridRegressionTechniqueforHouse PricesPrediction,”Proceedingsofthe2017IEEEIEEM-

978-1-5386-0948-4/17/$31.00 ©2017 IEEE, DOI: 10.1109/IEEN.2017.8289904

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