Shaker H. El-Sappagh, Samir El-Masri, A. M.

Riad, Mohammed Elmogy / International

Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
Data Mining and Knowledge Discovery: Applications,
Techniques, Challenges and Process Models in Healthcare
Shaker H. El-Sappagh1, Samir El-Masri2, A. M. Riad3, Mohammed
Elmogy4
1
Faculty of Computers & Information, Mansoura University, Egypt
2
College of Computer and Information Sciences, King Saud University, Saudi Arabia
3,4
Faculty of Computers & Information, Mansoura University, Egypt

ABSTRACT
Many healthcare leaders find themselves expenditures. As healthcare continues to become
overwhelmed with data, but lack the information more complex, the industry needs to find an
they need to make right decisions. Knowledge effective means of evaluating its large volume of
Discovery in Databases (KDD) can help clinical, financial, demographic and socioeconomic
organizations turn their data into information. data.
Organizations that take advantage of KDD Clinical decisions are often made based on
techniques will find that they can lower the doctors’ intuition and experience rather than on the
healthcare costs while improving healthcare knowledge rich data hidden in the database. This
quality by using fast and better clinical decision practice leads to unwanted biases, errors and
making. In this paper, a review study is done on excessive medical costs which affects the quality of
existing data mining and knowledge discovery service provided to patients. Integration of KDD
techniques, applications and process models that tools with EHR could reduce medical errors,
are applicable to healthcare environments. The enhance patient safety, decrease unwanted practice
challenges for applying data mining techniques variation, and improve patient outcome.
in healthcare environment will also be discussed. EHR is only a first step in capturing and utilizing
health-related data – the problem is turning that data
Key Words: Data mining, knowledge discovery, into useful information. Models produced via data
healthcare, Electronic health record, health mining and predictive analysis can form the
informatics. backbone of Clinical Decision Support Systems
(CDSS).
I. INTRODUCTION KDD(also known as knowledge extraction,
The advances in health informatics as data/pattern analysis, data archeology, data
Electronic Health Record (HER) make healthcare dredging, information harvesting, and business
organizations overwhelmed with data.There is a intelligence) is the extraction of interesting (non-
wealth of data available within the healthcare trivial, implicit, previously unknown and potentially
industry that would benefit from the application of useful) meaningful patterns or knowledge from huge
KDD tools and techniques. These techniques amount of data stored in multiple data sources such
transform the huge mounds of data into useful as file systems, databases, data warehouses and etc
information for decision making [3]. A proper by automatic or semi-automatic means [10].
medical database created with intention mining can KDD has evolved, and continues to evolve, from the
provide a useful resource for data mining and intersection of research fields such as machine
knowledge discovery [1]. The tools and techniques learning, pattern recognition, databases, statistics,
of KDD have achieved impressiveresults in other AI, knowledge acquisition for expert systems, data
industries, and healthcare needs to take advantage of visualization, and high-performance computing.
advances in this exciting field. A hospital typically More traditional query tools require the user to
has detailed data about every charge entered on a make many assumptions. For example, a user may
patient’s bill, which easily can reach hundreds of use a query tool to ask the question, “What is the
charges for only a few days’ stay. Each lab test, link between cholesterol and heart disease?” By
radiology procedure, medication, and so on is running this query, the analyst assumes that there is
recorded, whether in a clinical information system a relationship between cholesterol and heart disease.
or in a financial billing system. The power of KDD is that it will search the dataset
There is an enormous volume of data for all relationships, including those that may not
generated [8], but few tools exist in the healthcare have occurred to the analyst. With large datasets,
setting to analyze the data fully to determine the best there may be many variables interacting with one
practices and the most effective treatments. In another in very subtle ways. KDD can help find the
general, the healthcare industry lags far behind other hidden relationships and patterns within
industries in terms of information technology data.TheKDD is well-defined process consisting of

900 | P a g e
 Shaker H. El-Sappagh, Samir El-Masri, A. M. Riad, Mohammed Elmogy / International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
several distinct steps. Data mining is the core step, information that could be better used by employing
which results in the discovery of hidden but useful KDD techniques. Several examples include
knowledge from massive databases. This process identifying patients who should receive flu shots,
must have a model to control its execution steps. enroll in a disease management program and are not
KDD has many applications in healthcare in compliance with a treatment plan. Moreover,
such as patient diagnosis, patient treatment, historical data in EHR can help in management of
management of chronic diseases, prediction of chronic diseases and anticipating patient’s future
patients at risk for specific diseases, and in public behavior on the given history. EHR stores spatial
health.Because of the complexity of healthcare and demographic data which can help in public
environment, there are many challenges that face the health management and planning.
application and complete benefit from KDD. This - Finding themselves in an increasingly competitive
paper is organized as follows: section 2 discusses market, many healthcare organizations are now
data mining applications. Section 3 discusses KDD employing sophisticated marketing efforts. KDD
process models. Section 4 discusses data mining can help in this arena in ways similar to other
techniques. Section 5 is the data mining challenges, industries. Organizations can use their data to
and conclusion is discusses in section 6. identify those most likely to use their facilities and
the most effective marketing activities to reach those
II. DATA MINING APPLICATIONS IN individuals.
HEALTHCARE - KDD has proved as successful at identifying fraud
There is vast potential for data mining in the healthcare industry as it has in other
applications in healthcare. Generally, these can be industries. Several insurance companies use KDD
grouped as the evaluation of treatment effectiveness; techniques to sift through their claims, seeking to
management of healthcare; andCustomer identify fraudulent providers.
Relationship Management (CRM). More specialized - KDD is used to predict patient problems based on
medical data mining, such as analysis of DNA his medical history.
micro-arrays, lies outside the scope of this paper. - When researchers investigate the records of
- Evaluation of treatment effectiveness:Data mining patients with a particular disease to determine if
applications can be developed to evaluate the there are any risk factors in their histories that could
effectiveness of medical treatments (evidence based help predict the occurrence of the disease in other
medicine). By comparing and contrasting causes, patients; they may identify a new risk factor that
symptoms, and courses of treatments, data mining could help detect the disease sooner in other patients
can deliver an analysis of which courses of action and allow for more timely intervention.
prove effective such as predict optimum medication - Data mining techniques could be successfully
dosage. applied for detection and diagnosis with a
- Healthcare management: Data mining applications reasonably high performance of many medical
can be developed to better identify and track chronic diseases as cancer [4], diabetes [6, 24], kidney [7]
disease states and high-risk patients, design heart diseases [2], andsleep problems [5].
appropriate interventions, and reduce the number of - Hospital management for optimize allocation of
hospital admissions and claims. It can search for resources and assist in future planning for improved
patterns that might indicate an attack by bio- services. For example, forecasting patient volume,
terrorists.Moreover, this system can be used for ambulance run volume, etc and predicting length-of-
hospital infection control, or as an automated early- stay for incoming patients.
warning system in the event of epidemics. Accurate
prognosis and risk assessment as survival analysis III. DATA MINING PROCESS MODELS
for AIDS patients, predict pre-term birth risk, IN HEALTHCARE
determine cardiac surgical risk, predict ambulation Knowledge discovery is a process, and not
following spinal cord injury, and breast cancer a one-time response of the KDD system to a user's
prognosis. action. As any other process, it has its environment,
- CRM: Customer interactions may occur through its phases, and runs under certain assumptions and
call centers, physicians’ offices, billing departments, constraints. Figure 1 shows a typical decision
inpatient settings, and ambulatory care settings. It making environment. In the healthcare environment,
determines the preferences, usage patterns, and the source data in clinical databases and/or EHRs
current andfuture needs of individuals to improve can be queried directly using SQL. A Data
their level of satisfaction. Warehouse (DW) can be created to integrate data
from many sources and enhance data quality.
Other applications of KDD in healthcare are: Analysts can apply OLAP tools on the DW. Data
- Many providers are migrating toward the use HER warehouse is not enough for data analysis. Data
[12]. EHR store a large quantity of patient data on mining is required to discover hidden patterns in
test results, medications, prior diagnoses, and other either EHR or DW.
medical history. This is a valuable source of

901 | P a g e
 Shaker H. El-Sappagh, Samir El-Masri, A. M. Riad, Mohammed Elmogy / International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
KDD is an iterative or cyclic process that involves a [14]. The third model, which consists of eight steps,
number of stages. Although the specific techniques was developed byAnand& Buchner at about the
may vary from project to project, the basic process same time [15]. TheCRISP-DM (CRoss-Industry
is the same for all KDD problems. Standard Process for DM) process model that
includes six steps wasfirst proposed in early 1996
Simple SQL Query & Reporting [17] by a consortium of four companies: SPSS (a
Data SQL Query provider of commercialDM solutions), NCR (a
Sources Select
From
x
1
2
y z
database provider), Daimler Chrysler, and OHRA
Group By 3
(an insurancecompany). The last two companies
Data Warehouse Creation served as sources of data and case studies. Next is
Internal Data warehouse Slice & Dice Reporting the six-step process model of Cios et al. [16] by
Sources
(EHR) adopting the CRISP-DM model to the needs
ofacademic research community. The main
Knowledge Discovery extensions of the latter model include providing a
Data Selection Transformation Mining Knowledge
moregeneral, research-oriented description of the
External and Cleaning
Sources steps, introduction of several explicit
feedbackmechanisms and a modification of the
Figure 1:Decision Making Environment description of the last step, which emphasizes that
knowledgediscovered for a particular domain may
Following standard process in KDD help analysts by be applied in other domains. Another model is the
guiding them through analysis process exposing six-step KDD process model by Adriaans&Zantinge
those aspects that could otherwise be forgotten or (1996), which consists of DataSelection, Cleaning,
neglected. Figure 2shows the basic phases of the Enrichment, Coding, DM, and Reporting. Another
KDD process [11]. model is the four-step model by Berry &Linoff
(1997), which consists of Identifying the Problem,
Analyzingthe Problem, Taking Action, and
Measuring the Outcome. Another model is the five-
step SEMMA model by the SAS Institute Inc.
(1997), which consists of steps namedSample,
Explore, Modify, Model, and Assess. This model
was incorporated into commercial KDsoftware
platform SAS Enterprise Miner. Another model is
Figure 2: Phases in the KDD process
the seven-step model by Han &Kamber (2001),
Selection phase generates the target data set which consists of Learning the Applicationdomain,
from the database. Preprocessingsolve issues Creating a Target Data Set, Data Cleaning and
aboutnoise, incomplete and inconsistent data.The Preprocessing, Data Reduction andTransformation,
next phase is transformation of the preprocessed Choosing Functions of DM, Choosing the Mining
data into a suitable form for performing the desired Algorithm(s), DM, PatternEvaluation and
DM task. In the DM phase, a procedure is run that Knowledge Presentation, and Use of Discovered
executes the desired DM task and generates a set of Knowledge. Another model is the five-step model
patterns. However, not all of the patterns are useful. by Edelstein (2001), which consists of Identifying
The goal of interpreting and evaluating all the the Problem, Preparing theData, Building the
patterns discovered is to keep only those patterns Model, Using the Model, and Monitoring the
that are interesting and useful to the user and discard Model. Another model is the seven-step model by
the rest. Those patterns that remain represent the Klosgen&Zytkow (2002), which consist of
discovered knowledge. Definition and Analysis ofBusiness Problems,
This process is a time-consuming, Understanding and Preparation of Data, Setup of the
incremental, and iterative process by its very nature, Search for Knowledge,Search for Knowledge,
hence many repetition and feedback loops exists in Knowledge Refinement, Application of Knowledge
Figure 2. Individual phases can be repeated alone, in Solving theBusiness Problems, and Deployment
and the entire process is usually repeated for and Practical Evaluation of the Solutions. Finally is
different data sets. the seven-step model by Haglin et al. (2005), which
Many data mining process methodologies consists of Goal Identification, Target DataCreation,
are available. However, the various steps do not Data Preprocessing, Data Transformation, DM,
differ much from methodology to methodology. The Evaluation and Interpretation, andTake Action steps.
major KDD process models will be discussed
next:The first reported KDD model consists of nine A. NEW TRENDS IN KDD PROCESS
steps and was developed by Fayyad et al. in the mid- MODELS
1990s [10]. The next model, by Cabena et al., The future of KDD process models is in
consists of five steps and was introducedin 1998 achieving overall integration and distribution of the

902 | P a g e
 Shaker H. El-Sappagh, Samir El-Masri, A. M. Riad, Mohammed Elmogy / International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
entire process through the use of other popular examples where the values of the target attribute
industrial standards. Another currently very have been observed. The goal is to build a model
important issue is to provide interoperability and that can predict the value of the target attribute for
compatibility between different software systems new unseen examples. If the target value is nominal
and platforms, which also concerns KDD models. then the prediction task is known as classification
Such systems would serve end-users in automating, and each possible value for the target variable is
or more realistically semi-automating the KDD referred to as its class-label or class. For real-valued
systems. target attributes, the prediction task is known as
regression. Unsupervised learning refers to
A current goal is to enable users to carry modeling with an unknown target variable. In that
out KDD projects without possessing extensive case, models are solely descriptive. The goal of the
background knowledge, without manual data process is to build a model that describes interesting
manipulation, and without manual procedures to regularities in the data. Clustering is an example of
exchange data and knowledge between different a descriptive algorithm that is concerned with
DM methods. This requires the ability to store and partitioning the examples in similar subgroups.
exchange not only the data, but most importantly
knowledge that is expressed in terms of data models A. A GLANCE AT DATA MINING
generated by the KDD process and domain expert, TECHNIQUES
and meta-data that describes data and domain Data mining techniques can be classified
knowledge used in the process.The technologies, based on the database, the knowledge to be
which can help in achieving these goals, are XML discovered and the techniques to be utilized.
and Predictive Model Markup Language (PMML). Classification: if the target attribute is categorical
From the KDD point of view, XML is the key then the applicable techniques are decision tree
technology to [18]: induction, Bayesian classification, backpropagation
 Standardize communication between diverse data (neural network), based on concepts from
mining (DM) tools and databases. association rule mining, k-nearest neighbor, case
 Build standard data repositories sharing data based reasoning, genetic algorithms, rough set
between different DM tools that work on different theory, support vector machine and fuzzy set. If the
software platforms. target attribute is continuous then we use linear,
 Implement communication protocols between the multiple and non-linear regression.
DM tools. Clustering: many methods are applicable as
 Provide a framework for integration and partitioning methods (e.g. k-means, k-medoids),
communication between different DMKD steps. hierarchical methods (e.g. chameleon, CURE),
The information collected during domain and data density-based methods (e.g. DBSCAN, OPTIC),
understanding steps can be stored as XML grid-based methods (e.g. STING, CLIQUE) and
documents. Then, the information can be used for model-based methods (e.g. statistical and neural
data preparation and data mining steps, as a network approaches).
source of already-accessible information, cross- Association rules:There are many classifications of
platforms, and cross-tools. these algorithms such as according to the types of
values handled (e.g. Boolean and quantitative rules),
IV. DATA MINING TECHNIQUES IN according to the number of dimensions (e.g. single
HEALTHCARE and multi dimensional) and according to the level of
There are a variety of data mining abstraction (e.g. single and multi level rules).
techniques available, all with pros and cons, Graph mining, Social Network Analysis (SNA), link
depending on the business problem at hand and the mining and multi-relational data mining: these
data available for analysis. One of the most difficult techniques become increasingly important in
tasks is to choose the right data mining technique modeling sophisticated structures and their
which requires more and more expertise. The final interactions and relationships.
choice depends basically on:
- The main goal of the problem to be solved. Data mining algorithms are needed in
- The structure of the available data. almost every step in KDD process ranging from
Conceptual map ofmining technique helps in domain understanding to knowledge evaluation. For
selection for non-expert miners [26]. Discovered example, in data preprocessing, algorithms (e.g.
model (pattern, knowledge) from data mining genetic algorithms) are used in data integration,
algorithms can be predictive if they predict the value transformation and cleaning and in knowledge
of an attribute (referred to as the target attribute or discovery. Because of space restriction, we will not
target), by making use of the remaining attributes give examples of applicable techniques in healthcare
(referred to as predictive attributes). This is also real problems.
called supervised learning. The target attribute is the The most used software tools that provide these
focus of this process, and the data usually includes techniques are SPSS/ SPSS Clementine, Salford

903 | P a g e
 Shaker H. El-Sappagh, Samir El-Masri, A. M. Riad, Mohammed Elmogy / International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
Systems CART/MARS/TreeNet/RF, Yale (open history, social network of patients (Patient ID,
source), SAS / SAS Enterprise Miner, Angoss Mother ID, Father ID, and Family ID), and spatial
Knowledge Studio / Knowledge Seeker, KXEN, data as patient addresses. EHR may contain data
Weka (open source), R (open source), Microsoft with different formats as audio, video, image and
SQL Server, MATLAB and Oracle Data Miner. text data. These formats also need advanced data
mining techniques. The use of static techniques thus
V. DATA MINING CHALLENGES IN oversimplifies (or may hide) possible relationships
HEALTHCARE and thus support for longitudinal, temporal and
Application of KDD in healthcare faces many spatial semantics within the mining process is highly
challenges as [19, 20]: desirable. Episodic data is often the key to good data
1- Need for algorithms with very high accuracy mining. Techniques as anomaly detection [21],
because it is an issue of life or death. The problems difference detection [22], longitudinal X analysis
of missing, noisy and inconsistent data complicate [22] and temporal/spatial X analysis [23] are used.
this objective. Moreover, unlike other (arguably 7- Mining complex knowledge from complex data:
simpler) domains, the medical discipline itself is Using multi-relational data mining, mining
diverse, complex and, to an outsider, relatively knowledge in the form of graphs and mining non-
opaque. relational data as text and image are future
2- Active data mining: It needs two types of issues.Using text mining is challenging in analysis
triggers. First type used to fire data mining of physician free text describing patient diagnoses
technique to analyze the data automatically after and free text prescription. It can be used to
some time or making some updates. The second summarize patient conditions.
type is to enforce the discovered knowledge by 8- Utilizing data mining in CDSS creates a system
embedding the discovered knowledge within clinical of real-time data-driven clinical decision support, or
information systems. “adaptive decision support.” It can “learn” over
3- For the same problem, we need to apply many time, and can adapt to the variation seen in the
data mining techniques, compare their results and actual real-world population. The approach is two
select the most interesting one. pronged – developing new knowledge about
4- Automated or semi-automated KDD system: this effective clinical practices as well as modifying
challenge is critical since EHR data and medical existing knowledge and evidence-based models to
knowledge are changed continuously. KDD process fit real-world settings [9].
needs to start automatically or with a minor help 9- Most KDD systems today work with relational
from analyst. databases. KDD and data mining need to be
5- Previous (background) knowledge as concept extended to object-oriented and multimedia
hierarchy, domain expert knowledge, previously databases.
mined knowledge and rule template must be taken 10- Various security and privacy aspects: how to
in accountin data preparation, modeling and model ensure the users’ privacy while their data are being
evaluation phases [25]. Background knowledge can mined? One solution isAnonymousness and
be expressed in different formats: examples may be identification transformation. Besides data
found in the areas of decision rules, Bayesian preprocessing, in order to separate the relative
models, fuzzy sets and concept hierarchies. Figure 3 between patients and their records which may
shows that background knowledge must be taken in referenced some private information, anonym-
to account in the KDD process. oussness and identification transformation are also
5- The result of data mining system must be needed.
appended in the existing knowledge base 11- Distributed data mining, mining heterogeneous
(knowledge fusion). The new knowledge may and multi-agent data: many schemas can be utilized
update, remove, or add constraints to the existing as collecting data from all sites and then mine it, or
knowledge. mine the local data and distribute the discovered
knowledge. In distributed mining, one problem is
how to mine across multiple heterogeneous data
sources such as multi-database and multi-relational
mining.
12- Large databases: clinical databases are very
large and massive with hundreds of tables and
fields, millions of records. Using more efficient
Figure 3: Background knowledge role in KDD process algorithms, sampling, approximation, feature
6- Longitudinal, temporal and spatial support: There selection, and parallel processing can mitigate the
is a need for advancement in data mining techniques problem.
to deal with EHR environment [27]. EHR contains 13- High dimensionality: clinical environment has a
clinical data collected and summarized from many very large number of fields. It can be solved by
healthcare systems, temporal data forming patient using prior knowledge to know irrelevant variables.

904 | P a g e
 Shaker H. El-Sappagh, Samir El-Masri, A. M. Riad, Mohammed Elmogy / International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
14- Changing data and knowledge: Rapidly [2] K.Srinivas, Applications of Data Mining
changing data make previously discovered Techniques in Healthcare and Prediction of
knowledge invalid. Solving the problem of Heart Attacks, International Journal on
automatic KDD system can solve this problem. Computer Science and Engineering
15- Integration with other systems: A standalone (IJCSE), 02(02), 2010, 250-255.
discovery system may not be very useful. [3] M. Khajehei, F. Etemady, Data Mining and
Integration issues include integration with a DBMS Medical Research Studies, IEEE Second
(e.g. via a query interface), CPOE (Computer-based Int. Conference on Computational
Physician Order Entry system), spread sheets and Intelligence, Modeling and Simulation
visualization tools. (CIMSiM),2010, 119 – 122.
16- Distributed data mining: As EHR and clinical [4] L. Li, et al., Data mining techniques for
databases are mostly distributed. This situation cancer detection using serum proteomic
require enhancement to data mining algorithms and profiling, Artificial Intelligence in
KDD process to work with distributed data. Medicine, vol. 32(2), 2004, 71-83.
17- Developing a unified framework of data mining [5] O. Shmiel, et al., Data mining techniques
that can deal with the healthcare environment. for detection of sleep arousals, Journal of
18- Mining sequence data and time series data: A Neuroscience Methods, vol. 179(2), 2009,
particularly challenging problem is the noise in time 331-337.
series data. It is an important open issue to tackle [6] J. L. Breault,Data mining a diabetic data
19- Encoding problem: Healthcare data comprises warehouse, Journal of Artificial
both numeric and textual information for Intelligence in Medicine, 26 (1-2), 2002,
medications diagnostic tests, demographics, Pages 37-54.
problem lists, staff notes and images for radiology. [7] A. Kusiak, et al., Predicting survival time
Standard coding, such as International Classification for kidney dialysis patients: a data mining
of Disease (ICD-9-CM) codes for diagnoses and approach, Computers in Biology and
National Drug Codes (NDC) for drugs, SNOMED Medicine, vol. 35(4), 2005, 311-327.
CT (Systematized Nomenclature of Medicine - [8] N. Aggarwal, A. Kumar, H. Khatter, V.
Clinical Terms), Logical Observation Identifiers Aggarwal, Analysis the effect of data
Names and Codes (LOINC) and Classification of mining techniques on database, Advances
Interventions and Procedures (OPCS-4) was in Engineering Software 47(1), 2012, 164–
implemented when possible rather than having free 169.
text dataentry to facilitate research and analyses. [9] C. Bennett, T. Doub; Data Mining and
20- Collecting the default knowledge is a challenge. Electronic Health Records: Selecting
Default knowledge as “Only women can pregnancy” Optimal Clinical Treatments in Practice,
is used by domain expert or physician when making Proceedings of the 6th Int. Conference on
decision. This knowledge needs to be collected and Data Mining, pp. 313-318, 2011.
added in the data mining system’s knowledge base. [10] Fayyad, U., Piatetsky-Shapiro G. and
As a result, if a person is pregnant, then the system, Smyth, P., from data mining to knowledge
by default, knows that she is a woman. discovery: an Overview, In Fayyad, U,
Piatetsky-Shapiro, G, Smyth, P and
VI. CONCLUSION Uthurusamy, R (eds) Advances in
In this paper, we tried to make a review of Knowledge Discovery and Data Mining.
all KDD applications in healthcare. As the AAAI Press, 1996, 1-34.
healthcare is a complex environment, there are [11] U. Fayyad, P. Smyth, G. Piatetsky,
many useful and proved applications of data mining knowledge discovery and data mining
such as diagnoses, prognoses, treatment and others. towards a unifying framework, KDD-96,
The existing process models for data mining life American Association for Artificial
cycle are also reviewed. These models are critical Intelligence (AAAI), 1996.
for any KDD project because knowledge discovery [12] LI Yun, LI Xiang-sheng,The data mining
contains many phases. Existing data mining and knowledge discovery in
algorithms are also reviewed, and all problems, biomedicine,IEEE 5th Int. Conf.
challenges that face KDD in healthcare environment onComputer Science & EducationHefei
are collected. (ICCSE), China. 2010, 1050, 1052.
[13] J. Han, M. Kamber, Data Mining: Concepts
REFERENCES and Techniques, 2nd edition, Morgan
[1] S. K.Wasan, V. Bhatnagar, H. Kaur, The Kaufmann, 2006.
impactofdataminingtechniquesonmedical [14] Cabena, P, Hadjinian, P, Stadler, R,
diagnostics,in the proceeding of Data Verhees, J and Zanasi, A, Discovering Data
Science Journal, Vol. 5, 2006, 119-126. Mining: From Concepts to Implementation,
Prentice Hall. 1998.

905 | P a g e
 Shaker H. El-Sappagh, Samir El-Masri, A. M. Riad, Mohammed Elmogy / International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 3, May-Jun 2013, pp.900-906
[15] Anand, S, Buchner A; Decision Support [27] G. Hripcsak, C. Knirsch, L. Zhou, A.
Using Data Mining, Financial Time Wilcox, G. B. Melton, Bias Associated
Management, London. 1998. with Mining Electronic Health Records,
[16] Cios, K, Kurgan L.,Trends in data mining Journal of Biomedical Discovery and
and knowledge discovery,” In Pal, N and Collaboration, 6, 2011, 48‐52.
Jain, L (eds) Advanced Techniques in
Knowledge Discovery and Data Mining.
Springer, pp. 1–26. 2005.
[17] CRISP-DM, 2003, CRoss Industry
Standard Process for Data Mining,
http://www.crisp-dm.org.
[18] K. J. Cios, G. W. Moore, Uniqueness of
Medical Data Mining, Artificial
Intelligence in Medicine journal, 26(1-2),
2002, 1-24.
[19] Q. YANG, 10 Challenging problems in
data mining research, Int. Journal of
Information Technology & Decision
Making, 5(4)2006, 597–604.
[20] Ruben D. Canlas Jr., Data mining in
healthcare: Current applications and issues,
MSc. of Science in Information
Technology, Carnegie Mellon University,
Australia 2009.
[21] Y. Li, N. Wu, S. Jajodia, X.S.
Wang,Enhancing profiles for anomaly
detection using time granularities,” J. of
Computer Security, 10(1,2), 2002, 137–
157.
[22] Anthony K. H. Tung, Hongjun Lu, Jiawei
Han, Ling Feng, Efficient mining of
intertransaction association rules, IEEE
Transactions on Knowledge and Data
Engineering, 15(1), 2003, 43–56.
[23] C. Freksa, Temporal reasoning based on
semi-intervals, Artificial Intelligence, 54,
1992, 199–227.
[24] Noah Lee, A. F. Laine, J. Hu, F. Wang, J.
Sun, Sh.Ebadollahi, Mining electronic
medical records to explore the linkage
between healthcare resource utilization and
disease severity in diabetic patients, First
IEEE Int. Conf. on Healthcare Informatics,
Imaging and Systems Biology (HISB), on
page(s): 250 – 257, 2011.
[25] F. Alonso, L. Martínez, A. Pérez, J. P.
Valente,Cooperation between expert
knowledge and data mining discovered
knowledge: Lessons learned, ACM An
International Journal on Expert Systems
with Applications, 39(8), 2012, 7524–
7535.
[26] K. Gibert, M. Sànchez-Marrè, V. Codina,
Choosing the Right Data Mining
Technique: Classification of Methods and
Intelligent Recommendation Techniques,
International Environmental Modeling and
Software Society (iEMSs), www.iemss.org,
2010.

906 | P a g e