2025 年 16 巻 2 号 p. 97-106
Hydroponic greenhouses are a potential solution to the increasing demand for food and nutrition for the global human population. However, agriculture is influenced by complex relationships between multiple variables, particularly in regard to controlled environments, making it challenging for farmers to identify essential factors to create and manage optimal conditions for higher crop yields. Data-driven decision-making is an appealing solution for overcoming this challenge. The objective of this study was to use data science methods to identify the essential factors affecting a hydroponic tomato farm in Hino City, Tokyo, Japan. Specifically, this study identified the essential microclimatic and hydroponic factors that affect tomato yield. Further, this study compared the application of linear multiple regression and random forest regression models to identify the essential factors impacting the tomato harvest. Data sensors were installed in the greenhouse to obtain microclimatic and hydroponic data. Farm records, plant growth records, and tomato harvest data from three crop cycles (November 2021 to July 2024) were also collected. The moving average method was applied to smooth the data during preprocessing. The random forest regression model outperformed the linear regression model with a higher R2 value of 0.9, whereas the linear model had a lower R2 value of 0.31. Both models identified electrical conductivity supply, temperature, and the amount of water per plant as significant factors affecting tomato yield. Electrical conductivity showed a negative correlation, whereas temperature and the amount of water per plant showed a positive correlation, highlighting the importance of maintaining optimal levels for higher yields. This study provides practical insights into the essential yield-influencing factors and supports the implementation of customized management practices through data-driven decision-making, empowering smallholder hydroponic farmers to increase productivity.
