Web-based prediction of potential diabetes outbreaks using Django with the KNN algorithm

• Main Authors: Kusumaatmaja, Bintang Citra, Thamrin, Husni
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Data collection; Data processing; Diabetes; Diabetes mellitus; Euclidean geometry; K-nearest neighbors algorithm; Machine learning; Metabolic disorders
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0141932

Diabetes mellitus, also known as diabetes, is a chronic metabolic disease characterized by increasing sugar levels in the body exceeding the normal threshold. The number of diabetes cases in Indonesia according to the results of Riskedas based on parenchymal consensus in people aged over 15 years experienced a significant increase from 6.9% in 2013 to 8.5% in 2018. Therefore, it is necessary to have an early warning in the form of early predictions to prevent potential diabetes. The prediction is made in a website by utilizing the K-Nearest Neighbor (KNN) algorithm. In addition, this study also makes a comparison between the three types of distances, namely Euclidean, Manhattan, and Minkowski in that algorithm. The method used has several stages, namely data collection to collect data, data preprocessing to process raw data so that it can be processed by algorithms, data processing to carry out the algorithm learning process, model evaluation to analyze or evaluate models that have been made, and system development to create KNN algorithm user’s interface. This study found that the system can run well and the KNN algorithm able to predict with an accuracy of 71.4% using K=3 and the Euclidean distance calculation.