<![CDATA[management, including determining patient referral decisions at community healthcenters. However, these decisions often still depend on the subjective assessment ofmedical personnel, resulting in an inaccurate and ineffective process of identifyingdiabetes patient management. The purpose and objective of this research anddevelopment is to identify diabetes patient management for referral decisionrecommendations at Puskesmas using the K-Nearest Neighbor (KNN) approach toobtain a more accurate and effective process and results so that Puskesmas can morequickly provide appropriate follow-up based on patient laboratory test results. Thedata used in this study was diabetes patient data at Puskesmas, using variables suchas age, systolic and diastolic blood pressure, glucose tests, and referral to hospitals asthe target class. The results of the research and classification evaluation using theConfusion Matrix in KNN modeling based on this data showed that the number ofpatients included in TP=41, TN=38, FP=1, and FN=4, with an accuracy of 94.02%,precision of 97.62%, recall of 91.11%, and F1-Score of 94.25%. These values arecategorized as very good because they are able to predict classes correctly at themodeling stage. Thus, this study is considered feasible as a support for referraldecision recommendations in identifying the treatment of diabetic patients atPuskesmas]]>
Copyrights © 2025
