Sensor Malfunction Simulation and Data Imputation using Deep Learning in Precision Agriculture

Objectives: This research initiative develops a new GAN-based technique for handling sensor malfunction-induced data gaps in precision agriculture data through robust precision improvement of machine learning applications. Methods: A GAN-based imputation method operates to recover missing data points from agricultural datasets. The analysis of multiple sensor parameters created missing data by using conditional probability rules. The GAN received training through a combination of authentic data and simulated information to make its value predictions and imputation capabilities. Findings: The GAN-based imputation process proved better than standard methods since it achieved superior results in measurement accuracy and precision together with higher recall and F1-score. The proposed technique reduced errors related to missing data more significantly than the standard methods imputation. Traditional imputation techniques proved ineffective in datasets with sensor malfunctions, but the model achieved better outcomes in these situations. The experimental findings established that GAN-based imputation presents potential worth for real-time agricultural data processing because it helps produce reliable predictions which benefit precision farming operations. Novelty: GAN-based imputation method for agricultural IoT systems becomes a proposed solution which processes non-random missing data along with sensor malfunctions in an efficient light-weight system. Keywords: IoT, Precision Agriculture, Deep Learning, Machine Learning, Data Imputation, Sensors