Levenberg Marquardt Backpropagation Algorithm in Predicting

Potential Mortality in Heart Failure

Verdi Yasin

, Selli Oktaviani

, Muryan Awaludin

and Ifan Junaedi

Faculty of Computer Science, STMIK Jayakarta, Jakarta, Indonesia

Engineering Study Program, STIKOM Tunas Bangsa, Pemtangsiantar, Indonesia

Faculty of Industrial Technology, Universitas Dirgantara Marsekal Suryadarma, Jakarta, Indonesia

Keywords:

Levenberg Marquardt Backpropagation Algorithm, Predicting Potential Mortality, Heart Failure.

Abstract:

Heart failure is one of the most common disorders that attack the heart and blood vessels throughout the world,

resulting in a high average population death rate, and illness also has an impact ﬁnancially, especially for the

elderly. This study focuses on predicting the potential for death in heart failure using the Levenberg Marquart

algorithm. The data for predicting the potential for death in heart failure was taken from Kaggle, which con-

sisted of 299 records. Attributes used to predict potential death in heart failure consist of 11 attributes, namely

age, gender, anaemia, creatinine phosphokinase, diabetes, ejection fraction, high blood pressure, platelets,

serum creatinine, serum sodium, smoking, and death events. The results of this study are predictions of po-

tential death in heart failure with MSE training and testing = 0.0150. With 11-7-1 architecture.

1 INTRODUCTION

Artiﬁcial Intelligence (AI) is a general term that sug-

gests using computers to shape intelligent behaviour

with minimal human intervention. There are many

methods in AI, one of which is Artiﬁcial Neural Net-

work Backpropagation (Haring et al., 2022; Zuhruﬁl-

lah et al., 2022). Heart failure is a complex collection

of symptoms where a patient has various symptoms in

the form of shortness of breath at rest or when doing

activities accompanied by fatigue (Prihatiningsih and

Sudyasih, 2018). Backpropagation is a type of su-

pervised Artiﬁcial Neural Network training method.

This network architecture consists of an input layer,

a hidden layer, and an output layer (Sonang et al.,

2021). The Levenberg-Marquardt algorithm is a de-

velopment of the Backpropagation error algorithm.

This algorithm was built to overcome some of the de-

ﬁciencies that exist in the error algorithm (Andriani

et al., 2019; Maulana and Muslim, 2015; Sitompul,

2018).

2 RESEARCH METHODOLOGY

2.1 Methods of Data Collection

The method of data collection was carried out start-

ing from literature studies (sources used for various

purposes in the world of education both from scien-

tiﬁc journals and internet sources)(Abdullah and Han-

dayani, 2019; Lisa, 2015; Mokosuli et al., 2014).

2.2 Data Source

The data used in this study was taken from the

data website www.kagle.com, in the form of medi-

cal records of heart failure data consisting of 299 data

(Hikmayanti et al., 2014; Gultom et al., 2021; Ritha

et al., 2016). The data used for training is 100 records,

and the data used for testing is 100. The following are

299 attributes in the form of data that inﬂuence the

occurrence of potential death in heart failure (Yasin

and Junaedi, 2022; Zuhri, 2021).

Information:

X1 = Age

X2 = Gender

X3 = Anemia

X4 = Creatinine Phosphokinase

X5 = Diabetes

Yasin, V., Oktaviani, S., Awaludin, M. and Junaedi, I.

Levenberg Marquardt Backpropagation Algorithm in Predicting Potential Mortality in Heart Failure.

DOI: 10.5220/0012444100003848

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Advanced Information Scientiﬁc Development (ICAISD 2023), pages 72-75

ISBN: 978-989-758-678-1

Table 1: Heart Failure Medical Records Data.

No X1 X2 .. X10 X11 T

1 75 0 .. 1 0 1

2 55 0 .. 1 0 1

3 65 0 .. 1 1 1

4 50 1 .. 1 0 1

5 65 1 .. 0 0 1

6 90 1 .. 1 1 1

7 75 1 .. 1 0 1

8 60 1 .. 1 1 1

.. .. .. .. .. .. ..

299 50 0 .. 1 1 0

X6 = Ejection Fraction

X7 = High Blood Pressure

X8 = Platelet

X9 = Serum Creatinine

X10 = Serum Sodium

X11 = smoking

Target = Death Event

Heart failure medical record data is converted with the

following rules:

1. Age:

a. Toddler Age: 0–5 years = 0.1

b. Childhood: 5–11 years = 0.2

c. Early Adolescence: 12–16 years = 0.3

d. Late Adolescence: 17–25 years = 0.4

e. Early adulthood: 26–35 years = 0.5

f. Late adulthood: 36–45 years = 0.6

g. Early Old Age: 46–55 years = 0.7

h. Late Old Age: 56–65 years = 0.8

i. Age of Seniors: ¿ 65 years = 0.9

2. Anemia; if you have anemia = 1, otherwise = 0

3. Creatinine phosphokinase; if normal = 0 other-

wise = 1

4. Diabetes if you have a history of diabetes = 1; oth-

erwise = 0

5. Ejection fraction If normal = 0 otherwise = 1

6. High blood pressure If normal = 0 otherwise = 1

7. Platelets If normal = 0 otherwise = 1

8. Serum creatinine If normal = 0 otherwise = 1

The target of predicting the potential for death in

heart failure is if it causes death, then the value is

1; If not, then the value is 0. The results of con-

verting medical record data for heart failure can

be seen in Table 2.

Table 2: Heart Failure Medical Conversion Data.

No X1 X2 X3 .. X8 X11 T

1 0.9 1 0 .. 0 0 1

2 0.7 1 0 .. 0 0 1

3 0.8 1 0 .. 0 1 1

4 0.7 1 1 .. 0 0 1

5 0.8 0 1 .. 0 0 1

6 0.9 1 1 .. 0 1 1

7 0.9 1 1 .. 1 0 1

8 0.8 1 1 .. 1 1 1

.. .. .. .. .. .. .. ..

299 0.7 1 0 .. 0 1 0

2.3 Research Framework

In completing this research, the authors compiled the

research framework as follows:

Figure 1: Research framework.

2.4 Architectural Design

The architecture consists of 1 input layer block, one

hidden layer block and one output layer block. Figure

2, example of the 11-3-1 architecture .

3 RESULTS AND DISCUSSION

3.1 Best Training and Testing Results

Training data and predictive testing of potential death

from heart failure using the Mathlab 2011a applica-

tion with the Levenberg Marquardt method backprop-

agation algorithm. The best training and testing re-

sults are 11-7-1 with performance testing = 0.0150.

Levenberg Marquardt Backpropagation Algorithm in Predicting Potential Mortality in Heart Failure

Figure 2: Architectural Design.

3.2 Comparison of Training Results and

Testing of the Levenberg

Marquardt Method

After training and testing the Levenberg Markquard

method backpropagation algorithm with 11-2-1,11-

3-1,11-8-1,11-10-1,11-5-1,11-6-1,11 architecture -7-

1,11-4-1, the following is a comparison of the Leven-

berg Marquardt architecture with eight architectures.

Table 3: Training and Testing.

No Architecture Epoch Performance Performance

(iterations) Testing Training

1 11-2-1 454 0.0846 0.0846

2 11-3-1 1347 0.0780 0.0780

3 11-8-1 1174 0.0693 0.0693

4 11-10-1 126 0.0150 0.0150

5 11-10-1 126 0.0150 0.0150

6 11-5-1 514 0.0150 0.0150

7 11-6-1 96 0.0350 0.0350

8 11-7-1 59 0.0150 0.0150

9 11-4-1 100000 0.0426 0.0426

4 CONCLUSIONS

Based on the results and discussion described above,

it can be concluded that the Levenberg Marquart

backpropagation method can predict potential mor-

tality in heart failure with MSE training and testing

= 0.0150 with 11-7-1 architecture (Covid, 2022; So-

likhun et al., 2020a). Determination of the method

in backpropagation training is so inﬂuential on the re-

sults, and it’s just that the determination of the method

and pattern must be adjusted to the needs (Solikhun

et al., 2020b; Solikhun and Wahyudi, 2021).

REFERENCES

Abdullah, N. and Handayani, L. (2019). Peramalan

rate of return saham menggunakan metode brown’s

weighted exponential moving average dengan opti-

masi levenberg-marquardt: Forecasting the stock rate

of return using the brown’s weighted exponential

moving average method with optimization of leven.

Nat. Sci. J. Sci. Technol, 8(3):171–176,.

Andriani, Y., Wanto, A., and Handrizal, H. (2019). Jaringan

saraf tiruan dalam memprediksi produksi kelapa sawit

di pt. kre menggunakan algoritma levenberg mar-

quardt. Pros. Semin. Nas. Ris. Inf. Sci, 1(Septem-

ber):249,.

Covid, M. (2022). Analisis quantum perceptron un-

tuk memprediksi jumlah pengunjung ucokopi pe-

matangsiantar pada.

Gultom, W., Wanto, A., Gunawan, I., Lubis, M., and Ki-

rana, I. (2021). Application ofthe levenberg mar-

quardt method in predict the amount of criminality in

pematangsiantar city. J. Comput. Networks, Archit.

High-Performance Comput, 3(1):21–29,.

Haring, M., Grotli, E., Riemer-Sorensen, S., Seel, K., and

Hanssen, K. (2022). A levenberg-marquardt algorithm

for sparse identiﬁcation of dynamical systems. IEEE

Trans. Neural Networks Learn. Syst, pages 1–14,.

Hikmayanti, H., Kom, H., Komarudin, O., Si, S., and Kom,

M. (2014). Penggunaan algoritma backpropagation

levenberg marquardt dan teknik pengolahan citra dig-

ital untuk identiﬁkasi nominal uang kertas. J. Ilm. So-

lusi, 1(2):16–33,.

Lisa, Y. (2015). Levenberg marquardt dan regularisasi

bayes untuk prediksi curah hujan.

Maulana, M. and Muslim, M. (2015). Sistem prediksi tag-

ihan listrik usaha jasa laundry menggunakan jaringan

syaraf tiruan backpropagation. Unnes J. Math, 4(1).

Mokosuli, L., Weku, W., and Latumakulita, L. (2014).

Prediksi tingkat kriminalitas menggunakan jaringan

syaraf tiruan backpropagation: Algoritma levenberg

marquardt di kota manado berbasis sistem informasi

geograﬁ,”d’cartesian.

Prihatiningsih, D. and Sudyasih, T. (2018). Perawatan diri

pada pasien gagal jantung. J. Pendidik. Keperawatan

Indones, 4(2).

Ritha, N., Bettiza, M., and Dufan, A. (2016). Prediksi

curah hujan dengan menggunakan algoritma

levenberg-marquardt dan backpropagation,”j. Sustain,

5(2):11–16,.

Sitompul, H. (2018). Optimasi pemulusan eksponensial

dengan algoritma levenberg-marquardt hery andi sit-

ompul, s.si, m.si dosen kopertis wilayah i sumut, dpk

universitas darma agung.

Solikhun, L. and Wahyudi, M. (2021). Ann : Predicting

of state retail sukuk based on region in indonesia. J.

Phys. Conf. Ser, 1830(1).

ICAISD 2023 - International Conference on Advanced Information Scientiﬁc Development

Solikhun, M., Saﬁi, M., and Zarlis, M. (2020a). Back-

propagation network optimization using one step se-

cant (oss) algorithm. IOP Conf. Ser. Mater. Sci. Eng,

769(1).

Solikhun, M., Saﬁi, M., and Zarlis, M. (2020b). Resilient

algorithm in predicting fertilizer imports by major

countries. IOP Conf. Ser. Mater. Sci. Eng, 769(1).

Sonang, S., Deﬁt, S., and Ramadhan, M. (2021). Jepin (ju-

rnal edukasi dan penelitian informatika) analisis op-

timasi fungsi pelatihan machine learning neural net-

work dalam peramalan kemiskinan. Online]. Avail-

able:.

Yasin, V. and Junaedi, I. (2022). Algoritma backpropa-

gation metode powell-beale untuk prediksi penyakit

stroke.

Zuhri, A. (2021). Seminar nasional sains dan teknologi in-

formasi (sensasi) optimasi levenberg-marquardt back-

propagation dalam mempercepat pelatihan back-

propagation. Semin. Nas. Sains dan Teknol. Inf,

3(1):627–630,. Online]. Available:.

Zuhruﬁllah, I., Anggraini, F., and Dewantara, R.

(2022). Peramalan jumlah kasus baru hiv menu-

rut provinsi menggunakan machine learning dengan

teknik levenberg-marquardt. J. Comput. Syst. Infor-

matics, 3(4):212 221,.

Levenberg Marquardt Backpropagation Algorithm in Predicting Potential Mortality in Heart Failure