Use the methods of neural network, deep learning, and classification tree to predict the PM2.5 series of Beijing. You may use the same lagged explanatory variables as those of the Shanghai data in Example 4.3. For deep learning, classify the PM2.5 into six categories, namely

(0, 50], (50, 100], (100, 150], (150, 200], (200, 300], (300,∞].

Use the last 365 observations as the forecasting subsample.