ILinkNet-RNN Architecture-Based Predictive Data Analytics in Precision Agriculture for Crop Yield Prediction

Effective crop yield forecasting is vital for securing food supplies, maximizing resource efficiency, improving climate resilience, encouraging economic development, preventing losses, optimizing the supply chain and reducing risk in the agricultural sectors. This paper proposed an innovative predictive model in agriculture for crop yield prediction. Initially, the input data is preprocessed through improved normalization techniques to improve the stability of the data. The data consists of Crop, Crop-Year Season, Production, annual rainfall, fertilizers, and pesticides. The relevant features, namely, entropy-based features, Information gain, statistical features and Raw data from the preprocessed data, are extracted to enhance the prediction accuracy of the proposed model. Finally, a hybrid model, Improved LinkNet with Recurrent Neural Networks (IL-RNN), is proposed, which combines an improved LinkNet and Recurrent Neural Networks (RNN) model for the prediction process. In the Improved LinkNet model, an improved activation function is introduced to improve the overall performance and increase the convergence speed. Then, the prediction output from both models is fused through improved score-level fusion techniques for better-predicting results. Moreover, experimental investigation shows that the effectiveness of the proposed model is more accurate as compared to traditional prediction models with respect to various error metrics. Experimental results demonstrate that the Improved Link Net Recurrent Neural Network (IL-RNN) achieves a Median Absolute Error (MedAE) as low as 0.121 and a Root Mean Squared Error (RMSE) of 0.121 at 90% training data, outperforming other existing methods.