Implementation of Machine Learning Using the K-Nearest Neighbor Classification Model in Diagnosing Malnutrition in Children

The problem faced today is the lack of nutrition for children which causes stunting. One way to prevent stunting problems is to provide input to the community in Aceh for the importance of providing adequate nutrition for children. This study classifies toddlers who are identified as stunting with the K-NN model technology which is modeled in machine learning, the results are grouped. The purpose of this study was to determine the detection of malnutrition in toddlers and to classify data on malnutrition in toddlers using the k-means clustering method and the system that was built could be used as a reference to monitor the growth and development of children. Then in classifying malnutrition in children based on the results of the nutritional status criteria in toddlers, it can be known based on the index of Body Weight for Age (W/U), Height for Age (TB/U), and Weight for Height (W/TB). by entering data values from weight, height and gender of toddlers. The purpose of this study was to determine the detection of malnutrition under five at the Cut Meutia Hospital Kab. North Aceh. The process in the initial data analysis of Mr. ID, baby's name, gender, age, weight (kg), height (cm), the data to be classified for training data are 40 children in each region / village. In the assessment of nutritional status, it is classified as Malnutrition less than 3 SD or 70%, Malnutrition - 3 SD to < - 2 SD or 80%, Good Nutrition -2 SD to +2 SD, Over Nutrition >+2 SD. The results of the final score obtained are euclidean distance with a value of 1.3 with a ranking of malnutrition, age 1.6 months, weight (weight) 0.852, TB (height) 4.556 with euclidean distance with a value of 1.3 with a low ranking. For the second test data, age is 2.8 months, BB (weight) 0.222, TB (height) 4.556 with Euclidean distance with a value of 1.3 with a good rating of 0.778. The results of this study can be classified in children to children for each region in each region, village and sub-district of each Puskesmas in North Aceh Regency.


Introduction
Early childhood is the nation's successor in the field of education by being given good nutrition to improve quality (M.Garenne, 1992).The mortality rate of early childhood is around 12.4 million every year [2].Improving children's health is one of the government's important programs.During the growth period, children are susceptible to various diseases [2].Stunting is a nutritional imbalance that has an impact on decreasing the speed of infant growth [3].With a population density, things are very fast in getting problems about malnutrition and the dangers of stunting on the growth and development of early childhood [4].Stunting toddlers have difficulty in achieving their full growth and development potential, both physically and psychomotor.Stunting can have long-term consequences for children under the age of five, including impaired health, education, and productivity [5].
Problems that often occur in the nutritional status of children can be prevented in the form of counseling the community in fulfilling nutrition for toddlers so that stunting does not occur.The role of local governments, hospitals and health services is very much needed in seeing the magnitude of the nutritional status of children, especially in these areas [6].A standard measure known as a reference is often used in determining the classification of nutritional status itself [8].Furthermore, the fulfillment of vitamins plays an important role for the human body.Vitamins are useful for the process of growth, regulation, and improvement of body functions, while minerals play a role in several stages of energy metabolism reactions, growth, and maintenance of the body [8].
Nutritional status can be determined in several ways including anthropometry.The anthropometric guidelines for determining nutritional status are the parameters that are chosen and recommended, which include an assessment of age, sex, weight, and height.Body Mass Index (BMI) is recommended as a good indicator to determine nutritional status [9].Efforts to prevent stunting problems to the community, especially to mothers, by providing input, especially the people of North Aceh, on the importance of fulfilling nutrition for children to avoid stunting.This research is important to implement machine learning with a case based reasoning model in diagnosing malnutrition in children in seeing the grouping of toddlers who are identified as stunting or not by using expert system technology with the Case Based Reasoning model [10].
The attention of every parent is focused on the nutritional status and growth of their child, because the food consumed affects the development of the brain and memory.When a child's food intake is not sufficient to meet his nutritional needs, then the child is in danger of malnutrition.Stunting is described as a toddler who has a height lower than the standard for his age.Features indicate nutritional problems repeatedly and over a long period of time.Stunting affects school age, adolescents, and even adults [11].

Literature Review
An expert system is used to detect stunting in children. the results of the expert system are included in the Bayes Theorem classification to be able to draw conclusions from the choices obtained from data processing and instructions in detecting stunting early on.The results of this study also use the Bayes theorem to detect developmental delays in children, which can help people understand the symptoms and developmental delays in young children [12] [13].Data mining is widely used as an adequate and precise tool for predictors and diagnosis of many diseases.The technique used is competent to design a clinical support system because of its ability to discover hidden patterns and relationships in medical data [14].One of the most important uses is in the diagnosis of heart disease because it is one of the leading causes of death in the world [15].
Diabetes is a disease characterized by high blood sugar (glucose) levels.If diabetes is not well controlled.Tests with the right classification will be able to overcome the problems that will be measured in viewing the classification performance in managing datasets.The K-Nearest Neighbor (KNN) algorithm is a method for classifying objects based on the learning data closest to the object [16].then this study calculates calculating accuracy, precision, recall, and FMeasure based on the K value.The steps carried out in this research are splitting training data and testing data, applying the knn classification method, and calculating the performance of the method to be tested.Heart disease gets a lot of attention in medical research because of its great effect on the state of human health.
This study aims to compare the two algorithms and find out which algorithm can be used precisely in predicting the accuracy of heart disease data and is useful for considering health problems and being accurate as information.This study uses a comparison between the Supporting Vector Machine (SVM) algorithm and the K-Nearest Neighbor Algorithm (KNN).The method used is the classification method, one of the techniques in Machine Learning, to find parameters in linear equations that can be mapped input and output.The results show that the KNN algorithm shows an accuracy of 84.83% and 81.31% with normalization [17].
In the modeling classification with SVM (Support Vector Machine) image in the classification of sound types, the training data has each sound sample produced having its own energy value [18].Furthermore, the importance of capital that will be given to farmers is one of the causes of the decline in salt productivity so that the salt produced is not in accordance with market demand.The K-Nearest Neighbor classification model in the feasibility of providing capital uses a classification for salt farmer groups.The test result is a classification based on salt farmer groups with an accuracy value of 85, 45%.And there are results for grouping based on land suitability [19] [20].Collected early childhood disease data can assist medical personnel in diagnosing diseases that attack early childhood.This study aims to apply Principal Component Analysis (PCA) and K-Nearest Neighbor (K-NN) Classifier for the classification of early childhood diseases.The results of system evaluation using 150 test data show that the classification system by applying PCA and KNN Classifier has an accuracy value of 86% [21].

Data Collection
The data collection of this research was carried out in the following way: 1. Observation : is a method of collecting data by conducting direct observations to the object of research.2. Interview : This is a data collection technique by holding questions and answers or direct interviews with pediatricians at Cut Meutia Hospital.3. Library Study : Collecting data by studying problems related to objects and literature based data needed in research.

Research Stages
The research stages of Machine Learning Implementation with the K-Nearest Neighbor Classification are as follows:

Data Analysis
As for analyzing the data in the application of data mining, the Knowledge Discovery in Databases (KDD) stage consists of several stages, namely data selection, preprocessing, transformation, data mining, classification and evaluation.

Initial Data Analysis
The analysis of the initial data in the Implementation of Machine Learning With the K-Nearest Neighbor Classification Model in Diagnosing Malnutrition in Children is as follows: Tabel

Implementation of K-NN
The implementation of the KNN method for new cases is as follows:

Figure 1. The initial stage of implementing the KNN method
The results of the detection using the KNN method are as follows: The data clustering graph is as follows:

Figure 3 . 1
Figure 3. 1 Stages of Research 3.4 Data AnalysisAs for analyzing the data in the application of data mining, the Knowledge Discovery in Databases (KDD) stage consists of several stages, namely data selection, preprocessing, transformation, data mining, classification and evaluation.

Figure 2 .
Figure 2. Results of KNN and Cluster Models

5. 1 ConclusionFigure 3 .
Figure 3. Graph of Data Distribution The conclusions of the Implementation of Machine Learning With the K-Nearest Neighbor Classification Model in Diagnosing Malnutrition in Children are as follows: 1. Can find out the process of classifying children's malnutrition with the K-Nearest Neighbor Classification Model 2. Can assist the Health Service and Doctors in diagnosing malnutrition in children through the process of implementing machine learning in detecting child malnutrition 3.There is implementation of classification results in each cluster with TB 65, BB 18, TBN, 0.84, BBN, 0.64, distance 0.57 and the results of poor nutrition cluster, while the second test is with age 13, TB 70, BB 15 , BBN 0.68, BBN 0.73, distance 0.71 with good nutrition cluster results 5. Acknowledgement In further research, the training data in each area is more complete and the training data is more so that the more characteristics of the training data, the more correct the KNN classification and Clustering.

Table 3 .
1. Analisa Data Awal The nutritional status in the Implementation of Machine Learning With the K-Nearest Neighbor Classification Model in Diagnosing Malnutrition in Children is as follows: The calculation for the initial stage in the Implementation of Machine Learning With the K-Nearest Neighbor Classification Model in Diagnosing Malnutrition in Children is as follows: Calculation of the Initial stage