Applications of data augmentation in mineral prospectivity prediction based on convolutional neural networks