Evaluation of Key Parameters Using Deep Convolutional Neural Networks for Airborne Pollution (PM10) Prediction

Figure 17

Error distribution in every architecture. The order of the architectures is presented by order of appearance in this work. First is the multilayer perceptron (MLP), next is the one-dimensional CNN with one convolutional layer (1DCNN_1). The third model is the one-dimensional CNN with two convolutional layers (1DCNN_2). The fourth model is the two-dimensional CNN (2DCNN). The fifth model is the partial ensemble trained with the data of the month of January (PEJ). Finally, the sixth model is the final ensemble trained with the full dataset (FEJ).