Sistem Prediksi Pertumbuhan Badan Untuk Anak Di Indonesia Dengan Metode Artificial Neural Network

Lazuardi, Luthfi (2021) Sistem Prediksi Pertumbuhan Badan Untuk Anak Di Indonesia Dengan Metode Artificial Neural Network. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 05211740000072-Undergraduate_Thesis.pdf] Text
05211740000072-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (5MB) | Request a copy

Abstract

Pertumbuhan manusia ketika balita merupakan masa terpenting dalam hidup manusia, karena pada masa ini pertumbuhan manusia terjadi dengan pesat serta bisa dikatakan sebagai periode emas pertumbuhan. Namun sayangnya masalah pertumbuhan balita di Indonesia masih menjadi masalah tersendiri yang perlu ditangani. Hal ini dapat dilihat dari status gizi balita di Indonesia. Berdasarkan data status gizi Riset Kesehatan Dasar (Riskesdas) tahun 2018, proporsi balita di Indonesia dengan gizi buruk dan gizi kurang yaitu sebesar 17,7%, balita pendek dan sangat pendek (stunting) sebesar 30,8% serta proporsi balita kurus dan sangat kurus sebesar 10,2%. Proporsi balita gizi buruk dan kurang mengalami penurunan yang tidak signifikan dari tahun sebelumnya yaitu 17,8%. Sedangkan proporsi balita pendek dan sangat pendek serta balita kurus dan sangat kurus mengalami peningkatan dari tahun sebelumnya dengan proporsi 29,6% dan 9,5% berurutan. Permasalahan tersebut tentunya dapat dicegah, apabila masalah pertumbuhan pada balita dikemudian hari dapat diprediksi dengan akurat sebelumnya. Sudah ada penelitian-penelitian yang membahas terkait permasalahan ini seperti penelitian menentukan status gizi balita yang dilakukan menggunakan algoritma Fuzzy Tsukamoto, serta penelitian serupa dengan membandingkan dua metode yaitu metode k-Nearest Neighbor dan Artificial Neural Network di Probolinggo. Selain itu ada juga penelitian prediksi pertumbuhan anak di Bangladesh dengan menggunakan lima algoritma machine learning. Namun beberapa penelitian tersebut belum diterapkan di Indonesia. Selain itu upaya pemerintah saat ini yaitu menggunakan KMS masih sebatas mendeteksi masalah yang ada, belum bersifat memprediksi.
Oleh karena itu berdasarkan permasalahan yang ada dan penelitian sebelumnya tugas akhir ini menawarkan solusi dalam memprediksi pertumbuhan pada anak dengan memanfaatkan metode machine learning yaitu Artificial Neural Network (ANN), karena telah banyak penelitian sebelumnya yang menggunakan machine learning dalam memprediksi kasus-kasus tertentu di bidang gizi dan tumbuh kembang anak maupun di bidang kesehatan secara umum. Data yang digunakan yaitu data Indonesian Family Life Survey yaitu data survei rumah tangga dan komunitas yang terperinci di Indonesia yang dilakukan oleh RAND Coorporation, bekerja sama dengan lembaga penelitian di Indonesia.
Dari hasil analisis didapatkan bahwa model dengan 25 simpul pada hidden layer, menggunakan optimizer RMSprop dengan learning rate 0,01 dan dilatih dengan 3500 epoch dapat melakukan prediksi indeks massa tubuh (IMT) anak, namun dengan MAPE dan RMSE yang masih cukup besar dengan nilai masing-masing 10,27% dan 2,51 pada data pelatihan serta 11,45% dan 3,19 pada data pengujian. Hasil pengujian performa yang telah dilakukan menunjukkan bahwa seluruh fitur sistem prediksi dapat berjalan dengan baik, termasuk fitur utama yaitu melakukan prediksi IMT anak berdasarkan variabel-variabel tertentu menggunakan model yang diperoleh.
======================================================================================================
Human growth as a toddler is the most important period in human life, because at this time human growth occurs rapidly and can be said to be a golden period of growth. But unfortunately the problem of under-five growth in Indonesia is still a problem that needs to be addressed. This can be seen from the nutritional status of children under five in Indonesia. Based on data on the nutritional status of the Basic Health Research (Riskesdas) in 2018, the proportion of under-five children in Indonesia with severe malnutrition and malnutrition is 17.7%, short and very short (stunting) is 30.8% and the proportion of thin and very thin by 10.2%. The proportion of severe malnourished and malnourished in under-five children experienced an insignificant decrease from the previous year, namely 17.8%. Meanwhile, the proportion of short and very short under-five children and thin and very thin under-five children increased from the previous year with the proportions of 29.6% and 9.5% respectively. These problems can of course be prevented, if growth problems in under-five children in the future can be accurately predicted in advance. There are already studies that discuss this issue, such as the study of determining the nutritional status of toddlers using the Fuzzy Tsukamoto algorithm, and a similar studies comparing two methods, namely the k-Nearest Neighbor method and the Artificial Neural Network in Probolinggo. There is also a research on predicting child growth in Bangladesh using five machine learning algorithms. However, some of these studies have not been implemented in Indonesia. In addition, the current government's efforts to use Kartu Menuju Sehat (KMS) are still limited to detecting existing problems, not predicting it.
Therefore, based on the existing problems and previous studies, this final project offers a solution in predicting growth in children by utilizing the machine learning method, namely Artificial Neural Network (ANN), because there have been many previous studies that have used machine learning in predicting certain cases in the field of nutrition and child growth and development as well as in the health sector in general. The data used are Indonesian Family Life Survey data, a detailed household and community survey data in Indonesia conducted by the RAND Corporation, in collaboration with research institutions in Indonesia.
From the results of the analysis, it was found that the model with 25 nodes in the hidden layer, using the RMSprop optimizer with a learning rate of 0.01 and trained with 3500 epochs could predict a child's body mass index (BMI), but with MAPE and RMSE which were still quite large with a value of 10.27 % and 2.51 respectively on the training data and 11.45% and 3.19 respectively on the test data. The results of the performance testing show that all the features of the prediction system can work well, including the main feature, which is predicting a child's body mass index based on certain variables using the model obtained.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Pertumbuhan Balita, Artificial Neural Network, Sistem Prediksi, Indeks Massa Tubuh, Status Gizi Anak, Child Growth, Artificial Neural Network, Prediction System, Body Mass Index, Child Nutrition Status
Subjects: Q Science > Q Science (General) > Q325.5 Machine learning.
Q Science > Q Science (General) > Q325.78 Back propagation
Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science)
T Technology > T Technology (General)
T Technology > T Technology (General) > T57.5 Data Processing
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User: Luthfi Lazuardi
Date Deposited: 17 Aug 2021 04:13
Last Modified: 17 Aug 2021 04:13
URI: http://repository.its.ac.id/id/eprint/87455

Actions (login required)

View Item View Item