Spark Architecture and Ensemble-Based Feature Selection With Hybrid Optimisation Enabled Deep Long Short-Term Memory for Crop Yield Prediction

Precise prediction of crop yield is crucial for addressing the economic resilience and food security of agricultural countries. Current models for crop yield prediction struggle to fully understand the long-term trends and seasonal variations. Here, the Fractional Rider-Based Water Cycle Algorithm-Based Deep Long Short-Term Memory (FRWCA-DLSTM) is devised for crop production forecasting and addresses these issues. Primarily, the simulation of the IoT is performed. Then, the selection of Cluster Head (CH) and routing are done with the Rider-Based Water Cycle Optimisation (RWCO). Then, the crop production data are accumulated at the Base Station (BS), where Spark architecture is used for crop prediction. Here, the data partitioning is done using Deep Fuzzy Clustering (DFC). Next, the technical indicators are extracted. Then, the ensemble-based Feature selection is accomplished. Here, the ranking techniques are combined by a fusion function. The weight parameters are tuned by Hunter-Sparrow Search Optimisation (HSSO). Finally, the crop yield prediction is performed by DLSTM, which is trained using FRWCA. The FRWCA is developed by merging Fractional Calculus (FC) with RWCO. The performance of FRWCA-DLSTM shows the minimum mean absolute percentage error (MAPE), mean square error (MSE) and root mean square error (RMSE) of 0.103, 0.081 and 0.284, respectively.