Comparison of Decision Tree and K-Nearest Neighbors Methods on Classifying Household Electrical Appliances Based on Electricity Usage Profiles

Electricity use can be identified through commercial electricity meters, which are generally used today, where the information provided is only total electricity usage, which is less effective in electricity management. Electricity management can be done by monitoring and knowing active electrical appliances. In addition, load identification systems can be utilized in various applications such as electricity theft monitoring systems, electricity billing systems, Etc. This study designed a smart metering system to identify household electronic appliances based on their electricity usage profile. The contribution of this research is on how to implement a sensor and microcontroller to measure several electrical parameters consumed by household appliances and embed the system with K-Nearest Neighbors (K -NN) and Decision Tree (DT) algorithm for load classification. As the main contribution, the proposed method is to implement the proposed algorithm on an ARM-based processor and only send the result data as identified load and time stamp to the Internet. This approach will reduce the data size and energy consumption of smart devices for data transmission. The system was tested to classify some household electronic appliances (i.e., fans, televisions, smartphone chargers, rice cookers, and lamps), and both methods were compared under the same regulated conditions. The results show that the system can measure the electrical parameters of electronic appliances and identify the load type, with the DT’s prediction accuracy superior to K-NN in experiments under single-load and multi-load conditions.