A Data Factor Study for Machine Learning on Heterogenous Edge Computing

Abstract

As plants and animals grow, there are many factors that influence the changes that will affect how plants grow and how botanical experts distinguish them. The use of the Internet of Things (IoT) for data collection is an important part of smart agriculture. Many related studies have shown that remote data management and cloud computing make it possible and practical to monitor the functionality of IoT devices. In automated agriculture, machine learning intelligence is more necessary to use to automatically determine whether the correlation between learning factors influences plant growth patterns. In this research experiment, the relevant data are automatically collected through a detection device, and data modeling and computation are performed in an edge computing environment. At the same time, the data model is transmitted via the communication protocol, and another node is available for verification of the modeling and calculation results. The experimental results show that the single-point data-trained model is able to accurately predict the growth trend of the plants. In the case of verification of the second measurement point at a different space, the data model must be trained with more than two layers in order to improve the training results and reduce errors.