Authors:
Ryosuke Mizuno
1
;
Masaya Goto
1
and
Hiroshi Mineno
2
Affiliations:
1
Graduate School of Integrated Science and Technology, Shizuoka University, Japan
;
2
Graduate School of Integrated Science and Technology, Shizuoka University, Japan, Research Institute of Green Science and Technology, Shizuoka University, Japan
Keyword(s):
Imbalanced Data, Internet of Things (IoT), Artificial Intelligence (AI), Agriculture, Water-saving Cultivation.
Abstract:
Predicting the plant irrigation timing is an essential task in the domain of agriculture. A model that can predict the irrigation timing in tomato cultivation can assist new farmers who do not have sufficient experience and intuition. In this study, we propose an irrigation timing prediction method based on past irrigation data, environmental data, and plant water stress using a Random Forest model, which is a general machine learning method. Our proposed model reproduces irrigation decision making by an expert farmer for new farmers. Furthermore, we propose a method for resolving imbalances, focusing on the change in the characteristics of the state of plants due to irrigation. This is because irrigation timing data has a large imbalance, which is known to be difficult to formulate. Our proposed model clarifies the characteristics of the irrigation class, and can suppress its misjudgment. We evaluated the proposed method using tomato cultivation greenhouse data in Shizuoka, Japan. T
he results show a recall of 92% and f-measure 69% and hence, the irrigation timing can be predicted with high accuracy. In addition, the results show that the model works effectively to automatically determine the irrigation timing in greenhouse tomato cultivation.
(More)