This study develops a low-cost device for monitoring road roughness using the GY85 Sensor Module combined with the SIM7600CE-T Module. The device is mounted on a specialized motorcycle to collect data for machine learning modeling and result evaluation experiments. Several machine learning models are researched and evaluated for road classification results, including Support Vector Machine (SVM), Decision Tree, and K-Nearest Neighbors (KNN). The results show that the RBF kernel SVM model provides the best classification performance, with an accuracy of 95.75%. The experimental process indicates that the device operates stably, accurately, and has high mechanical durability. The classification results and some other information are displayed details on an app installed on a smartphone, while the data is stored on a memory card and a web server.