Statistical modeling for high-dimensional data and its application to E-agriculture

Reference No.	2022a008
Type/Category	Grant for General Research- Short-term Visiting Researcher
Title of Research Project	Statistical modeling for high-dimensional data and its application to E-agriculture
Principal Investigator	Hidetoshi Matsui（Faculty of Data Science, Shiga University・Associate Professor）
Research Period	September 12,2022. ～ September 16,2022.
Keyword(s) of Research Fields	Statistical modeling, Longitudinal data analysis, sparse regularization
Abstract for Research Report	We develop a statistical modeling method to describe the relationship between crop traits such as yield and quality and growing environmental factors such as temperature and solar radiation. The objective of this work is to elucidate the changes over the season in the relationship between environmental factors and the crop traits by analyzing data on crops that grow and yield every day over a long period of years. It is believed that the crop yields depend on the environmental factors over a period of days before the day of harvest. We will establish a statistical model and its estimation method that takes these relationships into account. With the recent development of measurement and measuring technologies, large-scale and complex structured data have been obtained in the field of agriculture. However, many farmers in the field are not fully utilizing these data, and the data remain accumulated. By utilizing these data, it is expected that it will be possible to predict the crop traits, and to give an guideline for optimal cultivation management to achieve the ideal amount of yield. In particular, we will clarify the period over which the environmental factors relate to the crop yield, depending on the season, by applying sparse regularization to a varying-coefficient functional regression model, one of the regression models in functional data analysis. From a methodological point of view, the novelty of this work is to develop an estimation procedure for the varying-coefficient functional linear model for the above purpose. Applying this method to the analysis of crop yield data, we can give farmers a guidance on when to begin cultivation management to achieve an ideal yield at a particular time of year.
Organizing Committee Members (Workshop) Participants (Short-term Joint Usage)	Hidetoshi Matsui（Faculty of Data Science・Associate Professor）