AI Service Development by Machine Learning using IoT Data and Simulation Data

Abstract

For estimation with AI from IoT data, a general method is to perform machine learning using IoT data as training data. Since IoT data also includes data related to transitional states, it is necessary that data cleansing to filter such unstable state data to make training data. It is an issue to reduce the number of trial and error of cleansing and machine learning. On the other hand, if information about the stable state of IoT data can be obtained from simulation, an estimation model can be created from such data, and then the data can be filtered to fit the estimation model. In this paper, we show that in the development of an AI service for estimating the amount of refrigerant for air conditioners, a highly accurate learning model can be created in a short period of time by performing machine learning using simulation data on stable conditions.