Development of software for analyzing and forecasting crop yields

V. K. Kalichkin¹, D. S. Fedorov^1,2, O. K. Alsova², K. Yu. Maksimovich^1
1Siberian Federal Research Center of Agricultural Biotechnology, Russian Academy of Sciences, pos. Krasnoobsk, Novosibirskii r-n, Novosibirskaya obl., 630501, Russian Federation
²Novosibirsk State Technical University, prosp. Karla Marksa, 20, Novosibirsk, 630092, Russian Federation

Abstract. The study aimed to develop software that implements the methodology (technology) for analyzing and predicting crop yields. It describes the algorithms, architecture, modules, and functions of the Crop Yield Analysis & Forecast (CYAF) software. The developed software consists of four modules: means of interaction with data, means of data presentation, data analysis, and forecasting. CYAF implements technology for studying long-term data on crop yields using a set of data mining methods, namely, methods of primary exploratory data analysis, variance data analysis, correlation data analysis, and decision tree. The input data for the software can be the results of field experiments conducted at any site with an unlimited number of fixed factors and conditions, the values of which can be measured both on qualitative and quantitative scales. The complexity of using data analysis methods and a combination of parametric and non-parametric approaches provide fairly high accuracy in predicting crop yields. The software is written in the statistical processing language R and the object-oriented programming language C#. The developed program is scalable. Its functions can be optimized and expanded by adding new machine learning algorithms. The software was tested on data on the yield of spring wheat in the northern forest-steppe of the Kuznetsk depression in Western Siberia. The determination coefficient of the predictive model is 0.83. The software has been successfully tested; all application modules work correctly in accordance with the declared functions; it is ready for implementation.

Keywords: productivity; forecasting; programming; data mining; R; C#; R.NET.

Author Details: V. K. Kalichkin, D. Sc. (Agr.), chief research fellow (e-mail: Этот адрес электронной почты защищён от спам-ботов. У вас должен быть включен JavaScript для просмотра.); D. S. Fedorov, leading programmer, master’s student; O. K. Alsova, Cand. Sc. (Tech.), assoc. prof.; K. Yu. Maksimovich, junior research fellow.

For citation: Kalichkin VK, Fedorov DS, Alsova OK, et al. [Development of software for analyzing and forecasting crop yields]. Dostizheniya nauki i tekhniki APK. 2021; 36(1):51-6. Russian. doi: 10.53859/02352451_2022_36_1_51.