U3895 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.

,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

NovelA Analyzation
Method for Predicting Health
ConditionBased on ECG Healthcare
Big Data using a Probabilistic Data
Compilation Method
Arun Kumar,M
Assistant Professor, Department IT, PSNA
College of Engineering and Technology,
Tamilnadu, India.
R. Vimala
Assistant Professor, Department of EEE, PSNA
College of Engineering and Technology,
Tamilnadu, India.
K.R. AravindBritoa
Assistant Professor, Department of EEE, PSNA
College of Engineering and Technology,
Tamilnadu, India.

Abstract: In the Healthcare management, a huge


volume of multi-organized patient data is created from
medical reports, doctor’s remarks and wearable body
sensors.The psychoanalysis of health care constraint and
prophesy of consequent physical conditions are still in the
instructive phase.A cloud facilitated giant data systematic
phase is the most excellent method to explore the
organized and unorganized data prepared from
healthcare management systems.In this paper, a
Probabilistic Data Compilation (PDC) Method is planned
and the link psychoanalysis of those composed data is
executed.Lastly a stochastic prophesy representation is
intended to predict the future health condition of the
most related patients based on their present health
conditions. Performance assessment of the planned
procedure is recognized through general model in the
cloud surroundings, that ensures 98% accurateness of
prediction and sustain 99% of CPU and bandwidth
consumption to decrease the analysis time.

Keywords: Huge data, Cloud, Healthcare, Prophesy,


Clustering and Classification.

Perspectivas em Ciencia da Informacao, v.22, sp.1, p1. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

1 Introduction
Wireless Sensor Networks(WSN’s) and mobile networks permitted in
hospitals and the outside patients are supervised through Internet Of
Things(IOT) [1] where patients are operated with various smart devices
like In-Plant Pacemaker,Electro Cardiogram(ECG),
Electromyography(EMG),Electro Encephalography (EEG) and Motion
Sensors. These wearable devices gathers health correlated data such as
blood pressure, heart rate and body temperature that would be useful in
physical condition tracking and medical healing. Huge data in
healthcare[2] is a systematic environment to grid the enormous quantity
of organized and unorganized patient data.According to the
psychoanalyst, the amount of data of USA healthcare system has attained
to 150 Exabyte’s in 2011[3] and has improved to zettabytesscale [4] in
the recent point of time.Also the California based health network Kaiser
Permanente has 9 million members and the data [3] composed from
Electronic Health Records [EHR’s] together with medical records,
pathological images doctor remarks vary from 26.5 to 44 peta bytes.

The health data are recognized has a huge data that is


distinguished by 5V’s in forms of Volume, Velocity, Value, Veracity and
Variety. The patient data are composed of peta or zeta bytes that
illustrate the volume.The velocity is stated in forms of data influx rate
from the patients.Variety clarifies the expanded data sets by means of
value to the organized, semi-organized and un-organized sets like medical
records,EHRs and radiological images and veracity explicate the reliability
of the data sets in accordance to data accessibility and validity.The
composed data are distorted into significant insights that explicate the
importance of 5V’s.Thus an appropriate raw data must be composed in a
well-organized manner in medical surroundings.In sophisticated
healthcare systems, the patients data are composed[5] through wearable
devices operated with special kinds of sensors.In recent times, the
development in mobile devices[6] like multi-censoredsmart phones are
also utilized as the data composed devices.Thus massive quantity of
patient data are produced inside a hospital network, that requires to be
accumulated and analyzed proficiently.So a cloud computing[7] permits
dispersed storage space and developed environment is necessary to
collect and process the healthcare data that could be accessed
everywhere and every time easily.

Currently different data intensive requests have come forward


that requires certain well-organized logical models. Several stochastic
approaches [8] are deemed by diverse authors in the current times for
health care structure analysis. Also, the similarity [9] among health
structure of a patient is regarded by the doctor for improved results. Huge
data logic is used in healthcare [10] to recognize the group of patient’s
illness and future forecast with the assist of several machine learning

Perspectivas em Ciencia da Informacao, v.22, sp.1, p2. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

devices [11]. The data’s are examined and utilized as insights, constantly
for patient concern in the learning healthcare system[12].

Yet, the existing design could not sustain both analysis and
procedure for huge quantity of multi-organized healthcare data.In recent
times, the larger presentation of cloud platform ensures a scalable and
distributable similar procedural framework(ie) MapReduce[13] for
healthcare data dispensation.Map Reduce has the capacity to process the
huge quantity of data in similar manner on a cloud.Thus a hybrid form of
stochastic similar processing structure is designed in medical
surroundings to process and examine the large quantity of healthcare
data.In this paper MapReduce similar processing structure[14] is used as
a backbone for healthcare huge data analysis.Additionally the PDC work is
expanded to prognosis a design for future health conditionprophesy of the
patients.

Related Works

The chronological and spatial connection of intellectual pilate exercise


data are examined in [15]. Still, partial works are carried out on
correlation analysis of healthcare parameters among various patients. In
hospital transience of critical unit, patients are forecasted using a confined
huge data-driven arbitrary forecast representation [16].A concise review
is executed based on benefits and difficulties of application and technical
necessities for hospital and BAN patient’s observation in [17]. Hu et al
[18] has projected the data acquirement process by using sensors, log
files and web crawler in different applications. Moreover the occurrences
of the patient visit are not deemed in data compilation.A methodology,
with unique huge data architecture for healthcare is proposed in [3] and
zhang et al [19], suggest a task-level adaptive and scalable MapReduce
structure, which could calculate the future influx rate of work load on the
map and reducedperiod.Likely MapReduce structure is designed to
decrease the re-calculation for increased iterative calculations in [20]. An
online society based health services is proposed in [21], in which the
health data are gathered and extracted through certain questionnaires
and their individual answers.A scalable and distributable process is
proposed in [22], to identify the similarity between patients by adjusting
the MapReduce structure. Moreover, the appointment regularity health
parameters and unknown symptoms of patients are most essential
although they are not taken into deliberation for evaluating and
processing the data in this work.

Future prediction of diseases is very decisive and significant for the


patients with chronic illness. In[23] various kinds of Artificial Neural
Network (ANN) methods are conversed for disease prophesy. Yet ANN,
takes more time for preparing the model due to expanded weights
correlated with every layer. Still, a few small variations in the input data
collection influence the model that provides unbalanced output. In [24],

Perspectivas em Ciencia da Informacao, v.22, sp.1, p3. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

the upcoming solidity is experimented by using the corrosion based


attribute graph for the medical prognosis in high-dimensional electronic
medical documents. An analytical structure is designed in [25] to
incorporate the HER data to efficiently predict the osteoporosis and bone
fractures. Henriques et.al [26] envisage the de-compensation of heart
malfunction by taking into consideration the physiological data of the
patients. Conversely the hidden symptoms of the sickness are not taken
into account in the present prophesy models.The bio-sensors [27] like
ECG,EMG and EEG are used to accumulate and transmit the health
constraints to backend servers for progression.Although various
researchers have projected the operation and sensing strategies of body
sensors for set data, none of them have enlarged the data set models of
inside and outside patients based on the occurrence of appointment to the
hospital. Moreover, the parallel analysis is integrated with sickness
prophesy between the patients in the hospital.

Thus the major perspective of this paper is to discover the eminent


characteristics of the diseases by establishing the correlation analysis of
healthcare parameters that are summarized below,

1) Projects a probabilistic data set model based on the occurrence of


out-patients appointment and number of data generated from the
patients with BAN.
2) Correlation analysis algorithms are proposed for the patients of
intra and inner sections of the hospitals.
3) An algorithm for foreseeing upcoming health condition of patients
based on their present health condition is planned.

Problem Definition
Let us consider a cloud based health care environment, with ‘h’
number of hospitals in a set as stated in figure1. Let
different departments be connected by one hospital and for simplicity, it is
implicated that similar and identical number of departments are present
in each hospital. Let be the collection of
number of departments connected with every hospital. Every
department nearby is combined with various number of doctors out-
patients and BAN patients, that are the basis for generating the huge
data. It is to be distinguished that out-patients are the patients who
attend a hospital for treatment without residing there whole night. BAN
patients are the chronic patients fixed with smart body sensors to
examine their health conditions regularly.
Let, represents number of doctors available in a set ,where
in the department of hospital and . Hence,
and . For example says
the doctor-2 from department 1 in hospital 5.
Let be the set of patients in which in the
department of hospital, and . Thus P, represents

Perspectivas em Ciencia da Informacao, v.22, sp.1, p4. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

number of patients present in the i th department of hospital.Thus,


, .For example, denotes the
patient-4, that is associated to the department -3 in hospital -2. It is
considered that patients with BANs are also hospitalized, that could be
neither a patient nor a BAN at a time. Likely, represents the number of
BANs appeared in a set , where , and
and various number of BAN are accessible in several departments within
the hospital.For example, denotes BAN-3 is associated to the
department-1 in hospital-2.

P1

Dp1
Pm
DC1
H1

P1
Gateway-1
Dpm
Pm

P1

Dp1 Gateway-1
Pm

H1
DCm
P1

Dpm
Pm

Figure-1: Proposed System Model

In the proposed PDC model, a window based chronological data set


and monitoring model is used to improve the quality of patient
monitoring. Let be a constant time frame that is partitioned
into number of windows, in which every window contains units of
time interval. Every time interval could be deemed as a minute, an hour,
a week, a month or a year based on the application going to be
handled.Consequently , , and denotes the amount of
data produced from BAN, doctors and patients correspondingly in every
window w. The gathered data inside window w are piled up in various

Perspectivas em Ciencia da Informacao, v.22, sp.1, p5. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

cloud data hub as shown in figure1. Let{ } be the


numbers of geo-dispersed data hub positioned in the cloud, where .
These data hub are joined through numbers of
gateways }where . In our structure,
represents number of those hospitals that are joined by those numbers
of geo-dispersed data hub via number of gateways.

Data Collection Model

In conventional healthcare systems, the patient data are gathered,


piled up and examined in a customary method that cannot maintain the
diagnosis of multifaceted health conditions.Moreover, in our PDC data set
scheme, patients, doctors and BANs are deemed in general as the basis of
data generated depending upon the rate of visits(r) of a patient rather
assuming only number of patients as examined in conventional
schemes.A window based[28] temporary data set and monitoring models
are utilized to augment the efficacy of patients monitoring.

110 90 120 150 130 140 190 140 100

1 2 3 4 5 6 t-2 t-1 t

w1 w2 w8
Figure-2: Data Collection in Windows Based Model

The window volume could be customized based on health


conditions. Consider this example, let A be the blood pressure data of a
patient, that is verified on the time framework , with
window frames }, as shown in figure2, where the interval
of every window holds 3units.In second window ( ), the normal blood
pressure data(Avg. ) of three time slots(4,5,6) is verified as 140, that is
higher than the vital condition (if(Avg. >125)). Those verified data by
time interval are gathered by using our proposed data acquirement
scheme and is conveyed to the data hub in the cloud for storage and
analysis.

In huge data healthcare surroundings the EHR,3Dimaging,


physiological data,medical,radiology images,genomic sequencing and
billing data are the basis of huge data which illustrate the volume.
Synchronized and urgent situation monitoring is the basis of streaming
data that illustrates the velocity of data. Although the majority of the
papers believe the patients physiological data as the huge data [3], we
incorporate the visiting rate of the patients to the hospitals in our huge
data processing models.

Perspectivas em Ciencia da Informacao, v.22, sp.1, p6. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

Data Acquisition Scheme

The number of appointments of a patient to seek advice from doctor


in various hospitals, required to be analyzed since they could produce
data during every appointment. Let us consider that a patient( ) makes
an appointment times to a department( ) within time
intervals.Every patient at a time of appointment let , , be the
probabilities of patients appointment rate of hospitals, departments and
doctors.To be considered that and are the smallest value of the
probability of a patient who makes an appointment in the hospital and
BAN correspondingly.
It is experimented that the possibility rises if the BANs
connected with several doctors and department increases. As stated in
[29], the medical test data and radio-logical images are deemed as
organized and unorganized collections of data correspondingly. In
accordance to the author’s one tera byte of medical text data and 19 tera
bytes of image data are produced by 25000 patients per year. In our data
acquisition scheme, text and image data of the patients are also deemed
that are produced during the appointment of the patients. Consider and
mega bytes, be the size of every text and image data correspondingly
and be the sum of data produced by a patient during a sole
appointment.Hence is the total sum of text( ) and image( ) data
of a patient which can be calculated in an efficient manner in
accordance to the number of department and number of hospitals. In
order to increase the efficiency of data retrieval and prediction on the
patient’s data an inter-cluster and intra-cluster methods are used. These
clustering methods give an arranged data in a uniform manner which
increases the speed of the searching. The prediction process look into the
patient’s information under a department in a hospital according to
association among the Identification key assigned to each elements. This
process is implemented and experimented in MATLAB software and the
results are given here. In order to verify the efficiency and performance of
the proposed approach the data is collected from cardiac disease patients
information with mild, moderate and sever attack information records
taken from All the data are collected fromCleveland and Hungarian clinic
[40] available for public use.

Experimental Results and Discussion

In the initial stage, the preprocessing method is applied to


normalize the raw data using some reserved functions. In accordance to
the number of patients the intra cluster method clusters the data in terms
of correlation. The results obtained from both existing and proposed are
shown in Figure-3. Instead of all the patients, 50 numbers of patients are
considered in each set to observe the data more clearly. The plot shows
that almost all the patients correlation values are greater than 0.95,
which leads to disease basedcorrelation exist between two sets in one

Perspectivas em Ciencia da Informacao, v.22, sp.1, p7. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

cluster. Therefore, the healthcare attributes of set 1 are highly correlated


with the attributes of set 2. From the figure, it seems that our proposed
algorithm is efficient for correlation analysis of the heart disease patients.

Figure-3: Correlation Obtained On Intra Cluster Method

In Figure-4, the inter cluster method is applied to cluster the


patients data. There are two different data sets are used in the
experiment. One is Cleveland and Hungarian with severe disease and the
other one is mild disease. From Figure-4, it is observed that some
correlation values are less than 0.6 and most of them are greater than
0.95. The values less than 0.6 are treated as the less correlated values,
where greater than 0.9 values are treated as highly correlated patients.
This correlation analysis could be extended to a classificationanalysis
based on the correlation values.

Figure-4: Correlation Obtained On Intra Cluster Method

Perspectivas em Ciencia da Informacao, v.22, sp.1, p8. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

Figure-5: Number of Parameters versus Variations in Correlation

The number of attributes shown in Figure-5 plays a vital role in the


correlation analysis. From the obtained simulation result in the beginning
stage of clustering process, the correlation analysis is varied due to the
less numbers of parameters. Even though this variation is minimized
during the number of parameters is increased. For both datasets the
correlation value is more drifted than the correlated values using both
clusters. From the result it is observed that the correlation values are
more stable according to large number of parameters.

Figure-6:Accuracy in terms of Number of Parameters

It is without a doubt observed in Figure-6 that the accuracy of


prediction is also elevated with increase in the quantity of attributes.
When best one characteristic is considered, the accuracy is touched
around 51%. However, it's far expanded up to 58% with attributes. The
accuracy is about eighty%, whilst the numbers of attributes are expanded

Perspectivas em Ciencia da Informacao, v.22, sp.1, p9. , Nov. 2017 ISSN 1413-9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

to 8 and eventually we got the maximum accuracy of 98%, when the


numbers of attributes are accelerated to 14.

Figure-7: CPU Time in terms of Increased Data Size

Processing time is a critical element to take a look at the performance of


the algorithms, which is proven in Figure-7. Here, the processing time is
described as the summation of challenge scheduling time with information
transfer time from distinct data centers to attain the records locality and
execution time. It is found that the processing time is decreased with the
aid of using a couple of virtual machines in distinctive records facilities.
The processing time of proposed set of rules is longer than the prevailing
algorithm as the responsibilities inside one statistics center await the
execution until the jogging responsibilities are finished. The proposed set
of rules with 3 facts centers is faster than the prevailing algorithm with 2
records centers due to parallel processing of the map obligations. Here,
the main advantage of using cloud platform is to lessen the processing
time.

Conclusion

In this paper a PDCmodel is intended for the cloud based healthcare


system. The existing and the PDC algorithms are structured for the intra
and inner cluster correlation analysis of the healthcare huge data. A PDC
algorithm is planned to prophesy the upcoming health status of the
patients, based on their present health condition with the precision of
98% cloud based MapReduce model is utilized as the procedural
framework for our huge data analysis.It is experimented that our
procedure could be used for several applications based on healthcare and
patient monitoring like heart sickness prediction, cancer severity
categorization.

Perspectivas em Ciencia da Informacao, v.22, sp.1, p10. , Nov. 2017 ISSN 1413-
9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

Our future work is to execute the proposed data analytic structure in


the actual healthcare field to examine the data in real-time data analytic
proposal like SPARK.

References
S. M. R. Islam, D. Kwak, M. H. Kabir, M. Hossain, and K.-S.Kwak, ``The
Internet of Things for health care: A comprehensive survey,'' IEEE Access,
vol. 3, pp. 678_708, 2015.

J. Andreu-Perez, C. C. Y. Poon, R. D. Merri_eld, S. T. C. Wong, and G.-


Z.Yang, ``Big data for health,'' IEEE J. Biomed. Health Inform., vol. 19,
no. 4, pp. 1193_1208, Jul. 2015.

W. Raghupathi and V. Raghupathi, ``Big data analytics in healthcare:


promise and potential,'' Health Inf. Sci. Syst., vol. 2, no. 1, pp. 1_10,
2014.

(Apr. 2014). The Digital Universe of Opportunities: Rich Data and the
Increasing Value of the Internet of Things. [Online]. Available: https://
www.emc.com/collateral/analyst-reports/idc-digital-universe-united-
states.pdf

K. Lin, F. Xia, W. Wang, D. Tian, and J. Song, ``System design for big
data application in emotion-aware healthcare,'' IEEE Access, vol. 4, pp.
6901_6909, 2016.

L. A. Tawalbeh, R. Mehmood, E. Benkhlifa, and H. Song, ``Mobile cloud


computing model and big data analysis for healthcare applications,'' IEEE
Access, vol. 4, pp. 6171_6180, 2016.

C. K. Dehury and P. K. Sahoo, ``Design and implementation of a novel


service management framework for IOT devices in cloud,'' J. Syst.
Softw.,vol. 119, pp. 149_161, Sep. 2016.

Z. Yu et al., ``Incremental semi-supervised clustering ensemble for high


dimensional data clustering,'' IEEE Trans. Knowl. Data Eng., vol. 28, no.
3, pp. 701_714, Mar. 2016.

M. Mukaka, ``A guide to appropriate use of correlation coef_cient in


medical research,'' Malawi Med. J., vol. 24, no. 3, pp. 69_71, 2012.

S. Rallapalli, R. R. Gondkar, and U. P. K. Ketavarapu, ``Impact of


processing and analyzing healthcare big data on cloud computing
environment by implementing hadoop cluster,'' ProcediaComput. Sci., vol.
85, pp. 16_22, May 2016.

S. Wang, X. Chang, X. Li, G. Long, L. Yao, and Q. Z. Sheng, ``Diagnosis


code assignment using sparsity-based disease correlation embedding,''

Perspectivas em Ciencia da Informacao, v.22, sp.1, p11. , Nov. 2017 ISSN 1413-
9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

IEEE Trans. Knowl. Data Eng., vol. 28, no. 12, pp. 3191_3202, Dec.
2016.

V. Tresp, J. M. Overhage, M. Bundschus, S. Rabizadeh, P. A. Fasching,


and S. Yu, ``Going digital: A survey on digitalization and large-scale data
analytics in healthcare,'' Proc. IEEE, vol. 104, no. 11, pp. 2180_2206,
Nov. 2016.

T. Huang, L. Lan, X. Fang, P. An, J. Min, and F. Wang, ``Promises and


challenges of big data computing in health sciences, '' Big Data Res., vol.
2, no. 1, pp. 2_11, 2015.

J. Dean and S. Ghemawat, ``MapReduce: Simplified data processing on


large clusters,'' Commun. ACM, vol. 51, no. 1, pp. 107_113, Jan. 2008.

C.-H. Wu and Y.-C. Tseng, ``Data compression by temporal and spatial


correlations in a body-area sensor network: A case study in Pilates motion
recognition,'' IEEE Trans. Mobile Comput., vol. 10, no. 10, pp.
1459_1472, Oct. 2011.

R. A. Taylor et al., ``Prediction of in-hospital mortality in emergency


department patients with sepsis: A local big data_driven, machine
learning approach,'' Acad. Emerg. Med., vol. 3, no. 23, pp. 269_278, Mar.
2016.

M. Patel and J.Wang, ``Applications, challenges, and prospective in


emerging body area networking technologies,'' IEEEWirelessCommun.,
vol. 17, no. 1, pp. 80_88, Feb. 2010.

H. Hu, Y. Wen, T.-S. Chua, and X. Li, ``Toward scalable systems for big
data analytics: A technology tutorial,'' IEEE Access, vol. 2, pp. 652_687,
2014.

F. Zhang, J. Cao, S. U. Khan, K. Li, and K. Hwang, ``A task-level


adaptive mapreduce framework for real-time streaming data in healthcare
applications,'' Future Generat.Comput.Syst., vols. 43_44, pp. 149_160,
Feb. 2015.

Y. Zhang, S. Chen, Q. Wang, and G. Yu, ``i2 MapReduce: Incremental


MapReduce for mining evolving big data,'' IEEE Trans. Knowl. Data Eng.,
vol. 27, no. 7, pp. 1906_1919, Jul. 2015.

L. Nie, M. Wang, L. Zhang, S. Yan, B. Zhang, and T. S. Chua, ``Disease


inference from health-related questions via sparse deep learning,'' IEEE
Trans. Knowl. Data Eng., vol. 27, no. 8, pp. 2107_2119, Aug. 2015.

M. Barkhordari and M. Niamanesh, ``ScaDiPaSi: An effective scalable and


distributable mapreduce-based method to _nd patient similarity on huge
healthcare networks,'' Big Data Res., vol. 2, no. 1, pp. 19_27, 2015.

Perspectivas em Ciencia da Informacao, v.22, sp.1, p12. , Nov. 2017 ISSN 1413-
9936
A Novel Analyzation Method for Predicting Health ConditionBased Arun Kumar,M et al.,
on ECG Healthcare Big Data using a Probabilistic Data Compilation Method

C.-H. Weng, T. C.-K. Huang, and R.-P. Han, ``Disease prediction with
different types of neural network classi_ers,'' Telematics Inform., vol. 33,
no. 2, pp. 277_292, 2016.

S. Gopakumar, T. Tran, T. D. Nguyen, D. Phung, and S. Venkatesh,


``Stabilizing high-dimensional prediction models using feature graphs,''
IEEE J. Biomed. Health Inform., vol. 19, no. 3, pp. 1044_1052, May 2015.

H. Li, X. Li, M. Ramanathan, and A. Zhang, ``Prediction and informative


risk factor selection of bone diseases,'' IEEE/ACM Trans. Comput. Biol.
Bioinf., vol. 12, no. 1, pp. 79_91, Jan./Feb. 2015.

J. Henriqueset al., ``Prediction of heart failure decompensating events


by trend analysis of telemonitoring data,'' IEEE J. Biomed. Health Inform.,
vol. 19, no. 5, pp. 1757_1769, Sep. 2015.

Perspectivas em Ciencia da Informacao, v.22, sp.1, p13. , Nov. 2017 ISSN 1413-
9936

You might also like