Method of corn yield prediction per grain applying Fuzzy Cognitive Maps

• Дата: 2024
• Автори: Popov, Mykhailo, Tarariko, Oleksandr, Alpert, Sofiia, Kokhan, Svitlana, Ilienko, Tetiana, Andreiev, Artem, Sybirtseva, Oksana
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: Scientific Centre for Aerospace Research of the Earth Institute of Geological Science National Academy of Sciences of Ukraine, Kyiv, Ukraine 2024
• Теми: Fuzzy Cognitive Maps; decision making; satellite and expert data; corn yield prediction
• Онлайн доступ: https://ujrs.org.ua/ujrs/article/view/261
• Назва журналу: Ukrainian Journal of Remote Sensing of the Earth

This work investigates the approach for predicting corn yield per grain using Fuzzy Cognitive Maps (FCMs) and an expert approach to describe the degree of influence of one factor (concept) on another. FCMs is a modeling methodology based on operating experience. It includes the main advantages of fuzzy logic and neural networks. FCMs represent a graphical model that consists of nodes-concepts which are connected with edges. Nodes-concepts describing elements of the system and edges represent the cause relationships among these concepts. FCMs can be applied in different areas especially for precision agriculture, yield modeling and yield prediction. FCMs can be also applied to model complex systems and can be applied for forecasting tasks. FCMs are ideal tool for modeling dynamic systems. The main advantages and specific features of the proposed algorithm are Ourflexibility, simplicity and high adaptability to different conditions. In this work, FCM approach was chosen to categorize yield in corn. This proposed methodology can apply satellite and expert data for yield prediction. This developed FCM model consists of nodes that represent the main concepts affecting yield, (such as potassium (K), humus, phosphorus (P), pH, nitrogen (N) and moisture contents, temperature, NDVI (Normalized Difference Vegetation Index), LAI (Leaf Area Index)). Potassium, P, pH, N and humus are expert data and temperature, moisture, NDVI and LAI are satellite data. Directed edges of FCMs show the cause-effect relationships between the concepts and yield. The main purpose of this study was to determine corn yield level using FCMs. Our model was applied for yield class prediction between three possible categories (low, middle and high) for three different experts. It was shown, that proposed algorithm can solve the problem of corn yield prediction. It should be noted that this algorithm can be applied for yield prediction of other agricultural crops.