[1] Data Mining Applications in Healthcare Sector a Study

8/20/2019 [1] Data Mining Applications in Healthcare Sector a Study

http://slidepdf.com/reader/full/1-data-mining-applications-in-healthcare-sector-a-study 1/7

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 2, ISSUE 10, OCTOBER 2013 ISSN 2277-8616

29IJSTR©2013www.ijstr.org

Data Mining Applications In Healthcare Sector: AStudy

M. Durairaj, V. Ranjani

ABSTRACT: In this paper, we have focused to compare a variety of techniques, approaches and different tools and its impact on the healthcare sectorThe goal of data mining application is to turn that data are facts, numbers, or text which can be processed by a computer into knowledge or informationThe main purpose of data mining application in healthcare systems is to develop an automated tool for identifying and disseminating relevant healthcareinformation. This paper aims to make a detailed study report of different types of data mining applications in the healthcare sector and to reduce thecomplexity of the study of the healthcare data transactions. Also presents a comparative study of different data mining applications, techniques anddifferent methodologies applied for extracting knowledge from database generated in the healthcare industry. Finally, the existing data miningtechniques with data mining algorithms and its application tools which are more valuable for healthcare services are discussed in detail.

Index Terms: Data Mining, Knowledge Discovery Database, In-Vitro Fertilization (IVF), Artificial Neural Network, WEKA, NCC2.————————————————————

1. INTRODUCTIONThe purpose of data mining is to extract useful informationfrom large databases or data warehouses. Data miningapplications are used for commercial and scientific sides[1]. This study mainly discusses the Data Miningapplications in the scientific side. Scientific data miningdistinguishes itself in the sense that the nature of thedatasets is often very different from traditional marketdriven data mining applications. In this work, a detailedsurvey is carried out on data mining applications in thehealthcare sector, types of data used and details of theinformation extracted. Data mining algorithms applied inhealthcare industry play a significant role in prediction anddiagnosis of the diseases. There are a large number ofdata mining applications are found in the medical relatedareas such as Medical device industry, PharmaceuticalIndustry and Hospital Management. To find the useful andhidden knowledge from the database is the purpose behindthe application of data mining. Popularly data mining calledknowledge discovery from the data. The knowledgediscovery is an interactive process, consisting bydeveloping an understanding of the application domain,selecting and creating a data set, preprocessing, datatransformation. Data Mining has been used in a variety ofapplications such as marketing, customer relationshipmanagement, engineering, and medicine analysis, expertprediction, web mining and mobile and mobile computing.

In health care institutions leak the appropriate informationsystems to produce reliable reports with respect to otherinformation in purely financial and volume relatedstatements. Data mining tools to answer the question thatraditionally was a time consuming and too complex to

resolve. They prepare databases for finding predictiveinformation. Data mining tasks are Association RulePatterns, Classification and Prediction, Clustering. Moscommon modeling objectives are classification andprediction. The reason that attracted a great deal oattention in information technology for the discovery ouseful information from large collections is due to theperception that we are data rich but information poor.Somethe sample data mining applications are:

Developing models to detect fraudulent phone ocredit-card activity

Predicting good and poor sales prospectus. Predicting whether a heart attack is likely to recu

among those with cardiac disease. Identifying factors that lead to defects in a

manufacturing process.

Expanding the health coverage to as many people aspossible, and providing financial assistance to help thosewith lower incomes purchase coverage [2]. Eliminatingcurrent health disparities would decrease the costsassociated with the increased disease burden borne bycertain population groups.Health administration ohealthcare administration is the field relating to leadershipmanagement, and administration of hospitals, hospitanetworks, and health care systems[1,3]. In the Healthcaresector Government spends more money.

Proposal in draft NHP 2001 is timely that Statehealth expenditures be raised to 7% by 2015 andto 8% of State budgets thereafter [21].

Health spending in India at 6% of GDP is amongthe highest levels estimated for developingcountries.

Public spending on health in India has itselfdeclined after liberalization from 1.3% of GDP in1990 to 0.9% in 1999. Central budget allocationsfor health have stagnated at 1.3%ofthe totaCentral budget. In the States it has declinedfrom7.0% to 5.5% of the State health budget.

_____________________________

M. Durairaj, Assistant Professor, Department ofComputer Science, Engineering and Technology,Bharathidasan University, India,Mobile No:+919487542202,(e-mail: [email protected] ).

V. Ranjani, Research Scholar, Department ofComputer Science, Engineering and Technology,Bharathidasan University, India,Mobile No: +918056930161,(e-mail: [email protected] ).

mailto:[email protected]












This paper mainly compares the data mining tools dealswith the health care problems. The comparative studycompares the accuracy level predicted by data miningapplications in healthcare. Infertility is on the rise across theglobe and it needs the sophisticated techniques andmethodologies to predict the end results of infertilitytreatments particulars IVF (in-vitro fertilization) treatments,since the cost of IVF procedure is on the rise. In this study,

we have taken this issue and compare the differenttechniques of data mining applications for predicting theSuccess rate of IVF treatment with the accuracy level. Thiscomparative study could be useful for aspiring researchersin the field of data mining by knowing which data miningtool gives an accuracy level in extracting information fromhealthcare data.

2. LITERATURE REVIEWA literature review is a text written by critical points ofcurrent knowledge including substantive find theoretical andmethodological contributions to a particular topic. Literaturereviews are secondary sources and do not report any newor original experimental work.

HianChyeKoh and Gerald Tan mainly discusses datamining and its applications with major areas like Treatmenteffectiveness, Management of healthcare, Detection offraud and abuse, Customer relationship management[1].

JayanthiRanjan presents how data mining discovers andextracts useful patterns of this large data to find observablepatterns. This paper demonstrates the ability of Datamining in improving the quality of the decision makingprocess in pharma industry. Issues in the pharma industryare adverse reactions to the drugs [2].

M. Durairaj, K. Meena illustrates a hybrid prediction system

consists of Rough Set Theory (RST) and Artificial NeuralNetwork (ANN) for dispensation medical data. The processof developing a new data mining technique and software toassist competent solutions for medical data analysis hasbeen explained. Propose a hybrid tool that incorporatesRST and ANN to make proficient data analysis andindicative predictions. The experiments onspermatologicaldata set for predicting excellence of animal semen iscarried out. The projected hybrid prediction system isapplied for pre-processing of medical database and to trainthe ANN for production prediction. The prediction accuracyis observed by comparing observed and predicted cleavagerate[20].

K. Srinivas, B. Kavitha Rani and Dr. A. Goverdhan discusses mainly examine the potential use of classificationbased data mining techniques such as Rule Based,Decision tree, Naïve Bayes and Artificial Neural Network tothe massive volume of healthcare data. Using an age, sex,blood pressure and blood sugar medical profiles it canpredict the likelihood of patients getting a heart disease[4].

ShwetaKharyadiscussed various data mining approachesthat have been utilized for breast cancer diagnosis andprognosis Decision tree is found to be the best predictorwith 93.62% Accuracy on benchmark dataset and also onSEER data set[5].

Elias Lemuye discussed the AIDS is the disease causedby HIV, which weakens the body’s immune system until itcan no longer fight off the simple infections that mosthealthy people’s immune system can resist. Aprioralgorithm is used to discover association rules. WEKA 3.6is used as the data mining tool to implement the AlgorithmsThe J48 classifier performs classification with 81.8%accuracy in predicting the HIV status[6].

Arvind Sharma and P.C. Gupta discussedData mining cancontribute with important benefits to the blood bank sectorJ48 algorithm and WEKA tool have been used for thecomplete research work. Classification rules performed welin the classification of blood donors, whose accuracy ratereached 89.9%[7].

3. DATA MININGData mining is the non trivial process of identifying validnovel, potentially useful, and ultimately understandablepatterns in data. With the widespread use of databases andthe explosive growth in their sizes, organizations are facedwith the problem of information overload. The problem o

effectively utilizing these massive volumes of data isbecoming a major problem or all enterprises.

DefinitionData mining or knowledge discovery in database, as it isalso known, is the non-trivial extraction of implicitpreviously unknown and potentially useful information fromthe data. This encompasses a number of technicaapproaches, such as clustering, data summarizationclassification, finding dependency networks, analyzingchanges, and detecting anomalies[8].

Development of data miningThe current evaluation of data mining functions and

products is the results of influence from many disciplinesincluding databases, information retrieval, statisticsalgorithms, and machine learning [9] (See Fig. 1).

Fig. 1. Historical perspective of data mining

History of Data Base and Data MiningData mining development and the history represented in theFig. 2. The data mining system started from the year o1960s and earlier. In this, the data mining is simply on fileprocessing. The next stage its Database managemenSystems to be started year of 1970s early to 1980s. In thisOLTP, Data modeling tools and Query processing areworked. From database management system there threebroad categories to be worked. First one is AdvancedDatabase Systems, this evaluated year of Mid-1980s topresent in this Data models and Application oriented

http://en.wikipedia.org/wiki/Secondary_sources

http://en.wikipedia.org/wiki/Secondary_sources





process are worked. The Second part is Data Warehousingand Data Mining worked since the year of the late 1980s topresent. The third part is Web based Database Systemsthis started from 1990s to present and in this Web miningand XML based database systems are included. Thesethree broad categories are joined and create the newprocess that’s called New generation of the IntegratedInformation system is started in 2000.

Fig. 2. History of Database Systems and Data Mining

Data Mining Application AreasData mining is driven in part by new applications whichrequire new capabilities that are not currently beingsupplied by today’s technology. These new applicationscan be naturally into two broad categories.

Business and E-Commerce Scientific, Engineering and Health Care Data

Data Mining TasksData mining tasks are mainly classified into two broadcategories:

Predictive model Descriptive model

Fig 3.3 Data mining models and tasks

4. DATA MINING APPLICATIONS INHEALTHCARE SECTORHealthcare industry today generates large amounts o

complex data about patients, hospital resources, diseasediagnosis, electronic patient records, medical devices etcLarger amounts of data are a key resource to be processedand analyzed for knowledge extraction that enables supporfor cost-savings and decision making. Data miningapplications in healthcare can be grouped as the evaluationinto broad categories[1,10],

Treatment effectivenessData mining applications can develop to evaluate theeffectiveness of medical treatments. Data mining candeliver an analysis of which course of action proveseffective by comparing and contrasting causes, symptomsand courses of treatments.

Healthcare management Data mining applications can be developed to better identifyand track chronic disease states and high-risk patientsdesign appropriate interventions, and reduce the number ohospital admissions and claims to aid healthcaremanagement. Data mining used to analyze massivevolumes of data and statistics to search for patterns thatmight indicate an attack by bio-terrorists.

Customer relationship managementCustomer relationship management is a core approach tomanaging interactions between commercial organizationstypically banks and retailers-and their customers, it is no

less important in a healthcare context. Customeinteractions may occur through call centers, physiciansoffices, billing departments, inpatient settings, andambulatory care settings.

Fraud and abuseDetect fraud and abuses establish norms and then identifyunusual or abnormal patterns of claims by physiciansclinics, or others attempt in data mining applications. Datamining applications fraud and abuse applications canhighlight inappropriate prescriptions or referrals andfraudulent insurance and medical claims.





Medical Device IndustryHealthcare system’s one important point is medical device.For best communication work this one is mostly used.Mobile communications and low-cost of wireless bio-sensors have paved the way for development of mobilehealthcare applications that supply a convenient, safe andconstant way of monitoring of vital signs of patients[11].Ubiquitous Data Stream Mining (UDM) techniques such as

light weight, one-pass data stream mining algorithms canperform real-time analysis on-board small/mobile deviceswhile considering available resources such as batterycharge and available memory.

Pharmaceutical IndustryThe technology is being used to help the pharmaceuticalfirms manage their inventories and to develop new productand services. A deep understanding of the knowledgehidden in the Pharma data is vital to a firm’s competitiveposition and organizational decision-making.

Hospital ManagementOrganizations including modern hospitals are capable of

generating and collecting a huge amount of data.Application of data mining to data stored in a hospitalinformation system in which temporal behavior of globalhospital activities is visualized[12]. Three layers of hospitalmanagement:

Services for hospital management Services for medical staff Services for patients

System BiologyBiological databases contain a wide variety of data types,often with rich relational structure.Consequently multi-relational data mining techniques are frequently applied to

biological data[13]. Systems biology is at least asdemanding as, and perhaps more demanding than, thegenomic challenge that has fired international science andgained public attention.

4. RESULTS OF COMPARATIVE STUDYThis chapter, a comparative study of data miningapplications in healthcare sector by different researchersgiven in detail. Mainly data mining tools are used to predictthe successful results from the data recorded on healthcareproblems. Different data mining tools are used to predict theaccuracy level in different healthcare problems. In thisstudy, the following list of medical problems has beenanalyzed and evaluated.

Heart Disease Cancer

HIV/AIDS Blood Brain Cancer Tuberculosis Diabetes Mellitus Kidney dialysis

Dengue IVF

Hepatitis C

In the Table 1,the most important healthcare problemsspecifically in disease side and research results have beenillustrated. The diseases are the most critical problems inhuman. To analyze the effectiveness of the data miningapplications for diagnosing the disease, the traditionamethods of mathematical / statistical applications are alsogiven and compared. Listed eleven problems are taken focomparison with this work.

TABLE 1. DATA MINING APPLICATIONS INHEALTHCARE

Graph chart formed by using this table with the values ohealth care problems, Data Mining tools and AccuracyLevel is as illustrated in Fig. 2. In this chart, the predictionaccuracy level of different data mining applications hasbeen compared.

Fig. 4. Chart for Accuracy Level of using Data mining toolsfor diagnosis





COMPARATIVE STUDY OF IVF SUCCESS RATEPREDICTIONThe section deals with the comparative study of threedifferent data mining application for predicting the successrate of IVF treatment. The process of data miningapplications, its advantages and results obtained arecompared. The detailed study of selected works gives abroad idea about the application of data mining techniques.

This study mainly compares the three different data miningapplications carried out on the prediction of the IVFtreatment success rate.

a) Application of rough set theory for medicalinformatics data analysis

The research work aims to analyze the medical data byapplying Rough Set Theory of data mining approach. Thedata reduction process has been done using rough settheory reduction algorithm. Rough set is mainly used toreduce the attributes without compromising its knowledgeof the original. To analyze the fertilization data, ROSETTAtool kit reduction algorithm is used in this work to producethe optimal reduct set without affecting the original

knowledge. The treatment success rate is predicted andtabulated as depicted in Table 2.

TABLE 2. IVF SUCCESS RATE PREDICTED BY ROUGHSET

The actual and desired outputs are compared with eachother. It also depicts that the success rate obtained afterreducing the number of attributes is 47%.

b)

Artificial neural network in classification andprediction

This research work is mainly aimed to predict and classifythe IVF treatment results using Artificial Neural Network(ANN). The artificial neural network is constructed withmulti-layer perception and back-propagation trainingalgorithm, and constructed network is trained, tested andvalidated using patients’ sample IVF data. This work finally

compares the success rate between desired output which isfield recorded data and actual output which is predictedoutput of neural network. In the Table 3, the comparisonbetween desired and actual output of the neural network isillustrated.

TABLE 3. IVF SUCCESS RATE PREDICTED BY ANN

This work finds the actual output using patients’ IVF data byapplying Artificial Neural Network. By comparing successrate, desired and actual output, the result obtained has aprediction accuracy of 73%.

c)

Modeling an integrated methodology of neuranetworks and rough sets for analyzing medicadata

This work is mainly aimed to develop a combined predictionsystem for analyzing medical data using Artificial NeuraNetwork and Rough Set Theory. Two kinds of rulesDeterministic and Non-deterministic are effected in theapplication of Rough set tool. For the rough set applicationthe software tool Neuro solution is used to predict theresult. The performance of the combined technique oArtificial neural network and rough set theory is described inthe Table 4.

TABLE 4. PERFORMANCE OF IVF SUCCESS RATE

PREDICTION USING HYBRID TECHNIQUE

The prediction accuracy of this hybrid approach ocombined use of ANN and RST is around 90%. Thesecomparison results of three different data miningapplications for predicting the success rate of IVFtreatments are shown in Table 5 and Fig. 5.





TABLE 5. COMPARISON BETWEEN THREEDIFFERENT DATA MINING APPLICATIONS

The application of combined Rough Set and Artificial NeuralNetwork yields better result when compared with othertechniques. It is observed that the hybrid technique ofcombined use of two or more machine learning tool yieldsbetter results than the use of a single technique for mininginformation from the database.

Fig. 5. The Success rate of Rough Set, ANN and HybridTechnique

5. CONCLUSIONThis paper aimed to compare the different data miningapplication in the healthcare sector for extracting usefulinformation. The prediction of diseases using Data Miningapplications is a challenging task but it drastically reducesthe human effort and increases the diagnostic accuracy.Developing efficient data mining tools for an applicationcould reduce the cost and time constraint in terms of humanresources and expertise. Exploring knowledge from themedical data is such a risk task as the data found are noisy,irrelevant and massive too. In this scenario, data miningtools come in handy in exploring of knowledge of themedical data and it is quite interesting. It is observed fromthis study that a combination of more than one data miningtechniques than a single technique for diagnosing orpredicting diseases in healthcare sector could yield morepromising results. The comparison study shows theinteresting results that data mining techniques in all thehealth care applications give a more encouraging level ofaccuracy like 97.77% for cancer prediction and around 70% for estimating the success rate of IVF treatment.

REFERENCES[1]. HianChyeKoh and Gerald Tan, ―Data Mining

Applications in Healthcare‖, journal of HealthcareInformation Management – Vol 19, No 2.

[2]. JayanthiRanjan, ―Applications of data miningtechniques in pharmaceutical industry‖, Journal oTheoretical and Applied Technology, (2007).

[3]. RubanD.Canlas Jr., MSIT., MBA , ― Data mining inHealthcare: Current applications and issues‖.

[4]. K. Srinivas , B. Kavitha Rani and Dr. A. Govrdhan―Applications of Data Mining Techniques inHealthcare and Prediction of Heart AttacksInternational Journal on Computer Science andEngineering (2010).

[5]. ShwetaKharya, ―Using Data Mining TechniquesForDiagnosis And Prognosis Of Cancer Disease‖International Journal of Computer ScienceEngineering and Information Technology

(IJCSEIT), Vol.2, No.2, April 2012.

[6]. EliasLemuye, ―Hiv Status Predictive ModelingUsing Data Mining Technology‖.

[7]. Arvind Sharma and P.C. Gupta ―Predicting theNumber of Blood Donors through their Age andBlood Group by using Data Mining ToolInternational Journal of Communication andComputer Technologies Volume 01 – No.6, Issue02 September 2012.

[8]. Arun K Punjari, ―Data Mining Techniques‖Universities (India) Press Private Limited, 2006.

[9]. Margaret H.Dunham, ―Data Mining Introductoryand Advanced Topics‖, Pearson Education(Singapore) Pte.Ltd.,India. 2005.

[10]. PrasannaDesikan, Kuo-Wei HsuJaideepSrivastava, ―Data Mining For HealthcareManagement‖, 2011SIAM International Conferenceon Data Mining, April, 2011.

[11]. Mobile Data Mining for Intelligent HealthcareSupport

[12]. ShusakuTsumoto and Shoji Hirano, ―Tempora

Data Mining in Hospital Information Systems‖.

[13]. David Page and Mark Craven, ―BiologicaApplications of MultiRelationalData Mining‖.

[14]. N. AdityaSundar, P. PushpaLatha and M. RamaChandra, ―Performance Analysis of ClassificationData Mining Techniques Over Heart Disease DataBase‖, International Journal of EngineeringScience & Advanced Technology, (2012).

[15]. HardikManiya, Mosin I. Hasan and Komal P. Patel―Comparative study of Naïve Bayes Classifier and




35IJSTR©2013ij

KNN for Tuberculosis‖, International Conference onWeb Services Computing (ICWSC) 2011Proceedings published by International Journal ofComputer Applications® (IJCA).

[16]. Andrew Kusiak , Bradley Dixonb and ShitalShaha,―Predicting survival time for kidney dialysispatients: a data mining approach‖, Computers in

Biology and Medicine 35 (2005) 311 –327.

[17]. B.Renuka Devi, Dr.K.NageswaraRao,Dr.S.PallamSetty and Dr.M.NagabhushanaRao,‖ Disaster Prediction System Using IBM SPSS DataMining Tool‖, International Journal of EngineeringTrends and Technology (IJETT) - Volume4 Issue8-August 2013ISSN: 2231.

[18]. Leah Passmore, Julie Goodside, Lutz Hamel,LilianaGonzalez, T Ali Silberstein And JamesTrimarchi, ―Assesing Decision Tree Models ForClinical In-Vitro Fertilization Data‖, TechnicalReport TR03-296

[19]. SaangyongUhmn, Dong-Hoi Kim , Jin Kim , SungWon Cho and Jae Youn Cheong, ―ChronicHepatitis Classification using SNP data and DataMining Techniques‖, Frontiers in the Convergenceof Bioscience and Information Technologies 2007.

[20]. M.Durairaj, K.Meena, ―A Hybrid Prediction SystemUsing Rough Sets and Artificial Neural Networks‖,International Journal Of Innovative Technology &Creative Engineering (ISSN: 2045-8711) VOL.1NO.7 JULY 2011.

[21]. R. Srinivasan, Health care in India – Vision 2020

Issues and Prospects.

AUTHOR'S PROFILE 1. He is currently working as a AssistantProfessor, Dept. of Computer ScienceEngineering &Technology, BharathidasanUniversity, Trichy, Tamilnadu, India. Hecompleted his Ph.D. in Computer Scienceas a full time research scholar a

Bharathidasan University on April, 2011. Prior to that, he

received master degree (M.C.A.) in 1997 and bachelordegree (B.Sc. in Computer Science) in 1993 fromBharathidasan University. Prior to this, he was working atNational Research Centre on Rapeseed-Mustard (IndianCouncil of Agricultural Research), Rajasthan, India, and aNational Institute of Animal Nutrition and Physiology(ICAR), Bangalore as a Technical Officer (ComputerScience) for 12 years. He has published 20 researchpapers in both national and international journals. Hisareas of interest include Artificial Neural Network, SofComputing, Rough Set Theory and Data Mining.

2. She received the Master Degree (M.C.Ain 2012 from Anna University, Chennai and

Bachelor degree (B.C.A) in 2009 fromBharathidasanUniversity, Trichy. She iscurrently pursuing asM.Phil ResearchScholar in the Department of ComputeScience, Engineering and Technology a

Bharathidasan University. Her areas of interest are DataMining, Artificial Neural Network, Rough Set Theory andNetwork.

Documents

[1] Data Mining Applications in Healthcare Sector a Study