Improving Air Pollution Prediction System through Multimodal Deep Learning Model Optimization

Many forms of air pollution increase as science and technology rapidly advance. In particular, fine dust harms the human body, causing or worsening heart lung-related diseases. this study, level in Seoul after 8 h is predicted to prevent health damage We construct a dataset by combining two modalities (i.e., numerical image data) for accurate prediction. addition, we propose multimodal deep learning model Long Short Term Memory (LSTM) Convolutional Neural Network (CNN). An LSTM AutoEncoder chosen time series data processing basic CNN. A Visual Geometry Group (VGGNet) (VGG16, VGG19) also CNN compare performance differences according network depth. The VGGNet standard architecture with multiple layers. Our using showed better than single only one modality (numerical data). Specifically, improved up 14.16% when VGG19 model, which has deeper network, was used rather VGG16 model.