Enhancing Crop Irrigation Prediction with Transformers Model in Smart Agriculture

Modem, Venkata Krishna Reddy (2025) Enhancing Crop Irrigation Prediction with Transformers Model in Smart Agriculture. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (687kB) | Preview
Preview	PDF (Configuration Manual) Download (569kB) | Preview

Abstract

Agriculture increasingly relies on intelligent technologies to overcome challenges such as water scarcity, climate variability, and inefficient manual irrigation practices. Accurate irrigation prediction is essential for optimizing crop growth, conserving water, and improving productivity in modern smart farming environments. However, existing machine learning and deep learning models often struggle to capture complex nonlinear feature interactions, depend on manual feature engineering, and lack robustness against noisy or inconsistent IoT sensor data, limiting their generalizability across diverse agricultural conditions. This study presents an intelligent irrigation prediction framework that integrates Machine Learning, Deep Neural Networks, and the Feature-Attentive Transformer (FT-Transformer), a transformer-based architecture tailored for tabular agricultural data. Using the CRISP-DM methodology, the research encompasses data preprocessing, exploratory analysis, feature encoding, model training, and comprehensive evaluation. The FT-Transformer employs feature tokenization and multi-head selfattention to learn relationships among key variables such as soil type, crop stage, moisture index, temperature, and humidity, offering a more adaptive modelling approach than conventional ML and DL methods. Experimental results show that while baseline ML and DL models deliver strong performance, the FT-Transformer consistently outperforms them across all metrics. Achieving an 0.9819 accuracy, precision, recall, and F1-score of 0.9819, it demonstrates exceptional predictive reliability and balanced classification. The findings confirm the FT-Transformer’s ability to capture intricate agricultural feature interactions and maintain robustness against data variability and sensor noise. Overall, the study establishes the FT-Transformer as a highly accurate, scalable, and effective solution for intelligent irrigation prediction in precision agriculture.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Horn, Christian UNSPECIFIED
Subjects:	H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Agriculture Industry H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Food Industry Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	08 Sep 2026 10:35
Last Modified:	08 Sep 2026 10:35
URI:	https://norma.ncirl.ie/id/eprint/9886

Actions (login required)

View Item