Unsupervised Anomaly Detection Using K-Means, Local Outlier Factor and One Class SVM

Abstract

Anomaly detection is one area that is still being researched today. An anomaly that occurs in data can be utilized in various ways, such as detection of money embezzlement, increasing production, improving the quality of data, preventing attacks on a network and others. This study aims to find anomalies in the data drug use in hospitals consisting of two datasets using the K-Means algorithm, Local Outlier Factor (LOF) and One Class Support Vector Machine (OC-SVM). The results of this study are that the three algorithms can find outliers. Based on its performance in both datasets, OC-SVM outperforms LOF and K-Means.