IoT Aware Piecewise Optimized Convolutional Deep Belief Network for Crop Yield Prediction in Cloud Environment

Objectives: To effectively enhance crop yield prediction accuracy and reduce the prediction time. To achieve accurate yield prediction results and minimize the error rate. Methods: A novel model called Piecewise Bayesian Optimized Convolutional Deep Belief Network (PIBO-CDBN) is proposed to deliver highly accurate predictions. It includes data harvesting, data augmentation, preprocessing, feature selection and classification. Performance of the PIBO-CDBN model is evaluated in python using Smart Farming Sensor Data for Yield Prediction dataset based on several key metrics with number of input samples. Findings: Results demonstrate that the PIBO-CDBN achieves superior accuracy by 96.44%, precision by 97.17% and recall 97.74% and F1-score by 97.45%. Also, the PIBO-CDBN reduces RMSE by 0.079, and MAPE by 0.0021 than traditional techniques. This optimized fine-tuning process improves the deep neural network's performance and overall learning accuracy in crop yield prediction tasks. Novelty: PIBO-CDBN utilizes Improved Convolutional deep belief network (CDBN) with several layers. Data preprocessing addressing missing values using imputation and eliminating outliers. Evaporative Cooling Relief mechanism identifies relevant features. Multivariate Bayesian linear regression analyzes the features and generates final yield predictions. During fine-tuning, error back-propagation algorithms adjust hyperparameters optimally using Adaptive Moment optimizer method. Keywords: Crop Yield Prediction in Cloud Environment, IoT, Piecewise interpolative robust Autoencoder model, Improved Convolutional deep belief network, Evaporative Cooling Relief mechanism