Indonesian Journal of Electrical Engineering and Computer Science Vol. 26, No. 3, June 2022, pp. 1556~1563 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v26.i3.pp1556-1563  1556 Journal homepage: http://ijeecs.iaescore.com Selecting the appropriate size of the graph for self-diagnostic model with graph density Sutat Gammanee 1 , Sunantha Sodsee 2 1 Department of Information Technology, Faculty of Information Technology and Digital Innovation, King Mongkut's University of Technology North Bangkok, Bangkok, Thailand 2 Department of Data Communication and Networking, Faculty of Information Technology and Digital Innovation, King Mongkut's University of Technology North Bangkok, Bangkok, Thailand Article Info ABSTRACT Article history: Received Jan 13, 2022 Revised Mar 21, 2022 Accepted Apr 1, 2022 Self-diagnosis is the concept of self-diagnosing disease from symptoms. We have the idea to create self-diagnostic models from diagnostic data. The data to be analyzed were from a medium-sized hospital in Thailand. The model is divided by structured data and unstructured data. The first step is to process structured data with cluster algorithms. The second step is to evaluate the unstructured data to group symptoms into a bipartite graph. After the graph was created, the model was divided into 10 levels, according to the level of similarity. This research aims to apply the concept of density graph, the Kappas and multiple line graph to selecting the appropriate diagnosis model. The results of all three experiments showed that the appropriate model was at a level of similarity at 40%. Keywords: Bipatite graph Graph density Machine learning This is an open access article under the CC BY-SA license. Corresponding Author: Sutat Gammanee Department of Information Technology, Faculty of Information Technology and Digital Innovation, King Mongkut's University of Technology North Bangkok Bangkok, Thailand Email: sutat@kru.ac.th 1. INTRODUCTION The current situation regarding the Coronavirus 2019 has made us realize that public health problems are an important problem of the country. Especially the problem of doctors, those are not sufficient to support the demand of the number of patients. Thailand has a ratio of 1 registered nurse per population of 412 people [1]. Therefore, the government has a policy that emphasizes the participation of the people in taking care of their own health [2], which diagnose themselves about the risks of different diseases to reduce the problem of seeing a doctor. Self-diagnosis is a concept of observing their own abnormalities, that reduce the risk of disease. Due to the current situation, social conditions, people work hard and they do not have time to go for a medical examination. People will go to the hospital only when they really get a serious health problem. The current situation makes the doctor's work is overload that there is no time to serve normal patients. Self-diagnosis is an option to reduce health problems and also choose to use public health services instead. Currently, the concept is applied machine learning to assist in self-diagnosis [3]-[6]. From the Figure 1, the researcher has the concept of self-diagnosis design. The concept has divided data into structured data and semi-structured data. Structured data include: preliminary values measured from body characteristics such as blood pressure values and heartbeat values, which is employed in screen patients