DLSTM with Adam Waterwheel Optimization for Groundwater Level Prediction in India

Accurate groundwater level prediction is vital for groundwater resource management along with risk assessment and land subsidence. Developing assistive techniques for predicting groundwater levels can assist in improving water resource management and planning. Hence, this work proposes a novel technique for groundwater level prediction in India based on deep learning. Here, at first, the input data is taken from the dataset, and data normalization is carried out using log scaling to standardize data and this is followed by feature selection, which is established using Harmonic mean distance. Finally, the groundwater level prediction is performed by Deep Long Short-Term Memory (DLSTM) trained with the proposed Adam Waterwheel Plant Algorithm (AWWPA). Here, AWWPA is established by the combination of Adam Optimization and Waterwheel Plant Algorithm (WWPA). Moreover, the proposed method is analyzed for its effectiveness by using metrics, like Mean-Squared Error (MSE), and Mean Absolute Error (MAE) Root-Mean-Squared Error (RMSE), and the analysis is effectuated based on different districts in India. The AWWPA-DLSTM is found to record a minimal MSE of 0.276, MAE of 0.305 and RMSE of 0.227.