Modelling variables affecting the yield of sugarcane fields using deep recurrent neural network

Document Type : Research Paper

Authors

• Hassan Zaki Dizaji ¹
• Kimia Shirini ²
• Adel Taheri hajivand ³
• nasim monjezi ⁴

¹ Biosystems Engineering Dept., Agricultural faculty, Shahid Chamran University of Ahvaz, Ahvaz, Iran,

² Department of Computer Engineering, Faculty of Electrical and Computer Engineering, Tabriz University, Tabriz, Iran

³ Department of Biosystem Engineering, Faculty of Agriculture, Tabriz University, Tabriz, Iran

⁴ Biosystems engineering Dept., Faculty of Agriculture, Shahid Chamran University of Ahvaz, Ahvaz, Iran

Abstract

Keywords

• Deep Learning
• Recurrent Neural Network with Long Short-Term Memory
• yield Prediction
• Sugarcane

Main Subjects

• ?????

EXTENDED ABSTRACT

Introduction

The yield optimization of sugarcane fields is a primary goal for agro-industry companies in Khuzestan, Iran. In recent years, advancements in data mining and machine learning have opened new avenues for enhancing agricultural productivity. This study investigates the application of deep learning techniques, particularly the Long Short-Term Memory (LSTM) recurrent neural network, to estimate sugarcane crop yields. By leveraging the capabilities of LSTM networks, the research aims to identify the most influential factors affecting farm performance and classify farms into three performance categories. This classification can provide valuable insights for improving management practices and achieving er yields. In this context, the application of deep learning algorithms, specifically LSTM recurrent neural network, presents a promising avenue for yield estimation. LSTM networks are particularly suited for time-series data, which is prevalent in agriculture due to the seasonal nature of farming activities and the longitudinal recording of environmental and operational parameters. By leveraging LSTM's ability to capture temporal dependencies and patterns, this study aims to develop a robust model for predicting sugarcane yields.

Materials and Methods

In this article, using one of the deep learning algorithms called deep recurrent neural network LSTM, the yield estimation of sugarcane crop has been done. In this research, by using LSTM deep recurrent neural network, the most influential features in the performance of farms are selected and based on them, the performance of farms is placed in one of three categories. In this regard, available data sets such as irrigation and drainage, soil and plant data were used to determine the effect of different combinations of these factors on production performance. This research is of analytical type and its database includes records of 1201 farms. The data required for this paper were obtained from Amir Kabir Sugarcane Agriculture and Industry Company during the crop years of 2014 to 2018. The analysis was done with the help of Python software. In this research, four algorithms of LSTM recurrent neural network, Multilayer Neural Network Perceptron (MLP), decision tree and Support Vector Machine (SVM) were used, and two-dimension reduction methods of Principal Component Analysis (PCA) and independent component analysis algorithm. (ICA) was also applied.

Results and Discussion

The results showed that executive and management indicators have an effect on changing the performance level of sugarcane fields. Also, crop variety and soil electrical conductivity have appeared as the most important independent variables in modeling in both algorithms; Therefore, the obtained results can help in planning and preparing optimal conditions to reach the set goals of the production rate. The results showed that the Long Short-Term Memory (LSTM) recurrent neural network algorithm performed better in the Principal Component Analysis (PCA) dimension reduction method. The values of R² equal to 0.97%, RMSE equal to 51.79, and RRMSE equal to 0.89 were obtained for this algorithm in the PCA method, which compared to the ICA method, which had values of R² equal to 0.91%, RMSE equal to 62.75, and RRMSE equal to 0.798. The results gave a better This shows that PCA had a better ability to reduce the dimensionality for this model.

Conclusion

This research demonstrates the potential of using LSTM deep recurrent neural networks for yield estimation in sugarcane farming. The accuracy achieved by the model underscores its applicability in real-world agricultural settings. Identifying crop variety and soil electrical conductivity as major influencers provides actionable insights for farm management. The study's results can assist agro-industry companies in Khuzestan in planning and implementing optimal conditions to achieve their production goals. Future research could expand on these findings by incorporating additional variables and exploring other advanced machine learning techniques to further enhance yield prediction accuracy.

Author Contributions

For research articles with several authors, a short paragraph specifying their individual contributions must be provided. The following statements should be used “Conceptualization, Kimia Shirini and Adel Taheri Hajivand; methodology, Kimia Shirini; software, Kimia Shirini.; validation, Hassan Zaki Dizaji, Adel Taheri Hajivand; formal analysis, Hassan Zaki Dizaji; investigation, Hassan Zaki Dizaji, Kimia Shirini; resources, Hassan Zaki Dizaji. data curation, Kimia Shirini; writing—original draft preparation, Hassan Zaki Dizaji, Kimia Shirini; writing—review and editing, Adel Taheri Hajivand, Hassan Zaki Dizaji and Nasim Monjezi; visualization, Hassan Zaki Dizaji; supervision, Hassan Zaki Dizaji; project administration, Hassan Zaki Dizaji; funding acquisition, Hassan Zaki Dizaji. All authors have read and agreed to the published version of the manuscript.” Please turn to the CRediT taxonomy for the term explanation. Authorship must be limited to those who have contributed substantially to the work re-ported.

Data Availability Statement

The data supporting the results reported in this study are available from the authors upon request. For inquiries, please contact the corresponding author.

Acknowledgements

The authors would like to appreciate the Vice Chancellor for Research and Technology of Shahid Chamran University of Ahvaz, Iran, for financial support under the special research grant number SCU.AA1402.505, and also to the R&D of Amir Kabir sugarcane Agro-Industry Company for preparing the data.

The authors would like to thank all participants of the present study.

Ethical considerations

The authors avoided data fabrication, falsification, plagiarism, and misconduct.

Conflict of interest

The author declares no conflict of interest.