As the demand for deploying Machine Learning (ML) and Deep Learning (DL) models at the network’s edge continues to grow, it becomes imperative to evaluate and compare the performance of various edge platforms. This study introduces an innovative edge-based pipeline for analyzing home energy and environmental data and addressing challenges in managing imbalanced label distributions. The comparison between random under-sampling and over-sampling strategies reveals the superiority of over-sampling, significantly improving accuracy and f1-score. More specifically, over-sampling using the Decision Tree (DT) technique enhanced accuracy and f1-score by 4.97% and 9.24%, respectively, when compared to the original results. The paper expands its contribution by conducting a thorough performance evaluation of various edge platforms, emphasizing critical metrics. The evaluated edge platforms include a Raspberry Pi 4b+, Jetson Nano Developer Kit, Odroid XU4, and Coral Dev Board. The study conducted an exhaustive comparison of the runtime of these prominent edge platforms when running ML/DL algorithms for tabular and image datasets during the testing phase. The Jetson Nano, particularly with its built-in GPU, demonstrates superior performance in ML/DL tasks, showcasing accelerated processing times for Convolutional Neural Network (CNN)-based algorithms. These findings emphasize the efficacy of the Jetson Nano in handling computationally intensive tasks for edge-based applications.