<![CDATA[Purpose - This study aims to develop an Internet of Things (IoT)-based Smart Energy System integrated with the Support Vector Machine (SVM) algorithm for real-time electricity consumption anomaly detection. The study addresses the increasing need for intelligent energy monitoring systems capable of identifying abnormal electricity consumption patterns efficiently and accurately in smart homes and smart building environments. Methods - This research employed a quantitative experimental approach using the Cross Industry Standard Process for Data Mining (CRISP-DM) framework. Electricity consumption data were collected using ESP32-based IoT devices integrated with ACS712 current sensors and ZMPT101B voltage sensors. Data preprocessing included cleaning, normalization using the Min-Max Scaling method, and anomaly labeling based on predefined energy consumption thresholds. The Support Vector Machine algorithm with a Radial Basis Function (RBF) kernel was implemented for anomaly classification. The dataset consisted of 25,000 electricity consumption records collected over 30 days, comprising 81.4% normal data and 18.6% anomaly data. Hyperparameter optimization was performed using Grid Search Cross Validation with 10-fold cross-validation. Findings - The experimental results demonstrate that the proposed SVM model achieved an accuracy of 96.80%, precision of 95.10%, recall of 94.90%, F1-score of 95.00%, and ROC-AUC of 97.20% in detecting electricity consumption anomalies. In addition, the MQTT-based IoT communication system achieved a data transmission success rate of 98.70% with low communication latency. Comparative evaluation results showed that the SVM algorithm outperformed Decision Tree, K-Nearest Neighbor, and Random Forest methods across all evaluation metrics. Implications - The proposed system contributes to the development of intelligent energy monitoring solutions capable of supporting real-time anomaly detection, reducing energy waste, and improving operational efficiency in smart energy environments. The lightweight computational characteristics of SVM also make the proposed approach suitable for IoT devices with limited computational resources. However, further testing using industrial-scale datasets is required to improve system generalization and large-scale deployment capability. Originality - This study presents an integrated smart energy architecture combining IoT-based real-time monitoring, MQTT communication, database systems, and SVM-based anomaly detection within a unified intelligent energy management framework.]]>
