Anomaly Detection on an IoT-Based Vaccine Storage Refrigerator Temperature Monitoring System

Maintaining temperature stability is paramount in research related to vaccine storage refrigerators. However, none have implemented a monitoring system with anomaly detection (AD) and alerts for anomalous temperatures in the vaccine storage refrigerators. The purpose of this study is to compare several AD methods to provide an optimum temperature alert system in an IoT-Based vaccine storage freezer temperature monitoring system. To implement the proposed system, an internet of things (IoT) architecture-based system is created with the message queue telemetry transport (MQTT) communication protocol and other specifications, such as a PT-100 sensor and a NodeMCU microcontroller. Based on the three AD methods applied and tested, histogram based outlier score (HBOS), minimum covariance determinant (MCD), and one class support vector machine (OCSVM), MCD has the best area under curve (AUC) score of 0.9999. Based on the value of sensitivity and specificity, MCD also has the most balanced value compared to other AD methods with values of 1 and 0.99, respectively. The contribution given by this research is an IoT system that can measure and monitor the temperature of the vaccine storage refrigerator and provide alerts if there are anomalies in the refrigerator temperature measurement.