Fine-Tuning Crop Classification: A Deep Dive into Hyperparameters for Long Short-Term Memory Networks

Remote sensing (RS) data and crop classification techniques provide useful information for crop yield estimation and prediction. Deep learning (DL) has seen a massive rise in popularity for remote sensing (RS)-based applications over the past few years. However, the performance of DL algorithms is dependent on the optimization of various hyperparameters since the hyperparameters have a huge impact on the performance of deep neural networks. The impact of hyperparameters on the accuracy and reliability of DL models is a significant area for investigation. The study region Charsadda is located in the Khyber Pakhtunkhwa, province of Pakistan. Five dates were chosen for satellite imagery in this investigation to capture the reflectance of crops at various growth stages. In this study, the grid Search algorithm is used for hyperparameters optimization of long short-term memory (LSTM) network for the RS-based classification. The hyperparameters considered for this study are optimizer, activation function, batch size, and the number of LSTM layers. In this study, over 1,000 hyperparameter sets are evaluated and the results of all the sets are analyzed to see the effects of various combinations of hyperparameters as well as the individual parameter effect on the performance of the LSTM model. The performance of the LSTM model is evaluated using the performance metric of minimum loss and average loss and it was found that classification can be highly affected by the choice of optimizer; however, other parameters such as the number of LSTM layers have less influence. This study shows that tuning the hyperparameters improves the model performance. The LSTM model for RS data yields the best performance with Adam, Nadam, RMSProp, and Adamax optimizers whereas it does not perform well with SGD, Adagrad, and Adadelta.