Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : March 04, 2024 - March 05, 2024
In recent years, automated machine learning (AutoML) has been developed with democratizing machine learning. AutoML tools can generate classifiers and prediction equations with a few clicks by simply collecting data and setting up the desired problem, and their use is expanding. On the other hand, classifiers and prediction formulas generated by AutoML are not necessarily easy to implement in edge devices such as IoT sensors. In this study, we implement several machine learning models generated by AutoML on IoT sensors and discuss how to realize the problem of identifying anomalies in the machine structure. We focused on a simplified discrimination method of deterioration of a water pipe and its implementation with an IoT sensor module by machine learning. We used DataRobot, an AutoML platform, to generate 66 models to determine the deterioration of water pipes. Furthermore, we considered the discrimination accuracy of models based on decision boundary of machine learning algorithms. The results show that the logistic regression has best accuracy, followed by the decision tree algorithms which have good accuracy.
