description.txt

We developed an AI model designed to predict soil moisture levels using a Long Short-Term Memory (LSTM) neural network. The process began with data acquisition and preprocessing, where we gathered soil moisture data, geographical coordinates (latitude and longitude), soil composition metrics (clay, sand, silt content), and auxiliary soil moisture data (sm_aux) from various NASA datasets.
Our model is an LSTM neural network. We structured the network with two LSTM layers and incorporated dropout layers to prevent overfitting. The model was trained using the Adam optimizer and mean squared error as the loss function. We implemented early stopping to halt training when the validation loss ceased improving.
Finally, we validated our model by generating predictions for specific locations and dates, outputting the results to CSV files for easy interpretation and further use and analysis.