Spiralizer Strategy with Wolf Pack Algorithm and Temporal Convolution Network – Gated Recurrent Unit for Coconut Crop Yield Prediction

The estimating coconut crop yield across large areas is crucial for ensuring food security and early to predicted. However, current spatial heterogeneity learning methods face limitations, such as selecting irrelevant features across regions and difficulties in achieve better performance in crop yields. In this research, Spiralizer Strategy with Wolf Pack Algorithm (SWPA) for feature selection and Temporal Convolution Network – Gated Recurrent Unit (GRU) for classification combined to performed efficiently and achieve better performance. The SWPA efficiently select relevant feature using a spiral search patter, leading to refined feature set that is fed into classification model. The TCN is designed to learn sequential data by applying convolutional layer that maintain the order of sequence. The GRU technique specifically designed to handling sequential data and long term dependencies, reducing the risk of overfitting. Initially,data acquisition and pre-processing involved Min–Max normalization, label encoder which efficiently scaling and categorized values into 0 and 1. The proposed method achiever better outcomes such as better MAE of 1.50, MAPE of 8.05, MSE of 7.23, RMSE of 0.35, \({R}^{2}\) $R^2$ of 0.654, respectively. The existing method such as Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) are compared to proposed method.