Smart Net-Zero

IEEE Internet of Things Journal | Mar 15, 2023

A recent study conducted by National Yang Ming Chiao Tung University in Taiwan focused on the Bacillus bacteria, which is widely used in the agricultural biotechnology industry to enhance crop growth. Traditionally, studies on Bacillus analysis were performed in laboratories due to the difficulty of conducting them in open field farming. The researchers aimed to predict the amount of Bacillus using innovative IoT (Internet of Things) and machine learning technologies, and they developed a method called AgriTalk.

The challenge was that only a small dataset was available for training the AI model, as soil analysis for Bacillus is time-consuming and limited. By using just five data items per farm, the researchers trained the AgriTalk system to predict the Bacillus levels for the following four months. The results were promising, with the mean absolute percentage errors (MAPEs) ranging from 6.73% to 19.76%.

The predictions obtained through AgriTalk provide valuable information for fertilization management in agriculture. Interestingly, the system showed higher accuracy in farms covered by peanut shells (average MAPE of 13.24%) compared to those covered by rice husks (average MAPE of 15.43%). These findings contribute to more efficient and effective agricultural practices.


AgriTalk architecture