An innovative approach to predict atmospheric rivers: Exploring convolutional autoencoder