Deep Learning IoT Enabled Driven Data Centric Crop Yield Forecasting Using Optimized Temporal-CRN Networks with BIANO Algorithm

Read full text Summarise with VIRAT
Authors: Nirmala Teegala Papaiah Teegala, Prakash Ganapathy, Anbu Karuppusamy Shanmugam
Cited by: 0
References: 21

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

An IoT-enabled deep learning framework is presented for reliable data-driven crop yield forecasting using real-time agricultural sensor information.The proposed approach integrates a Temporal Attention-Enhanced Convolutional Recurrent Network with hierarchical feature learning to effectively model temporal dependencies, spatial correlations, and nonlinear growth dynamics in agricultural data streams.Multi-source IoT data acquired from soil, weather, and crop condition sensors are subjected to systematic preprocessing, including noise removal, normalization, standardization, and dimensionality reduction using an Autoencoder-based Feature Compression combined with Independent Component Analysis (ICA) to preserve salient agronomic patterns while minimizing redundancy.The refined dataset is partitioned into training and testing subsets following an 80:20 ratio to ensure unbiased model validation.Feature extraction is performed through statistical and higher-order statistical descriptors, augmented with two novel adaptive features-adaptive weighted skewness and adaptive weighted kurtosis-to enhance sensitivity toward yield-critical variations.To further improve predictive efficiency, a Bio-Inspired Adaptive Nutrient Optimization algorithm is employed for optimal feature subset selection and dynamic hyperparameter adjustment, enabling improved convergence and generalization of the TA-CRN model.The proposed framework is implemented using Python 3.7.9 and evaluated as a multi-class yield categorization task, where continuous harvested yield values (kg/ha) are discretized into four agronomically meaningful classes-Low Yield, Medium Yield, High Yield, and Very High Yield -using training-set-derived percentile thresholds; performance is then assessed using accuracy, precision, sensitivity, specificity, F-measure, and macro-averaged classification statistics.Comparative experiments conducted against conventional deep learning models such as CNN, RNN, LSTM, and DNN demonstrate the superiority of the proposed approach.Experimental results indicate that the proposed TA-CRN with the BIANO framework achieves a maximum prediction accuracy of 96.5%, validating its effectiveness for precision agriculture and intelligent crop yield forecasting.

Fields of Study & Subjects

Smart Agriculture and AI Remote Sensing in Agriculture Internet of Things and AI

Publication details

Journal / source
Traitement du signal
Volume
43
Issue
01
Pages
199–222
Publication date
28 February 2026
Language
EN