python == 3.8.5

Data

• Delhi

  • AQI

    • IAQI for one/several pollutants, taken from monitoring stations. OpenAQ - Script to download
      • Data till July 2019 is problamatic, it mostly has data from a single station
    • Location coordinates (lat, lon).
    • Meteorological features: weather, temperature, pressure, humidity, wind speed and direction. Hourly meteorological data
      • ERA5 hourly data-single levels resolution is too low for Delhi data. 37 stations share only 4 ERA5 locations - notebook link
      • ERA5 hourly data-Land - 37 stations share 13 ERA5 locations. "0.1" degree resolution in lat-long.
      • Checking datasets on CCAI website resources - link

• Beijing

  • AQI & Meteorological

    • U-Air - Author profile - Dataset Link - Download
    • ADAIN - Same author as above - Dataset download
    • KDD Cup 2018 - Dataset page

Results

Bejing dataset from U-Air : spatial interpolation for each time-stamp.

• rf - RandomForest
• svr - Support Vector Regression
• gp_m32/12 - Gaussian Process regression with Matern32/12 kernel
• elst - ElasticNet
• dt - Decision Tree
• dkl - Deep Kernel Learning
• gp_sm_gpytorch - Spectral Mixture kernel known for extrapolation in GPyTorch
• gp_rbf_torch - My own implementation of GP in PyTorch
• mlp - neural network trained on each time-stamp
• mlp_gen - neural network trained on all time-stamps

RMSE

	fold_0	fold_1	fold_2	fold_3	fold_4	fold_5	average
rf	39.11	24.69	28.87	33.87	29.10	23.60	29.87
svr	43.37	26.48	28.52	33.61	29.76	25.96	31.28
gp_m32	42.15	27.69	30.74	53.32	28.65	22.54	34.18
gp_m12	41.39	27.01	29.62	35.31	28.98	22.59	30.82
gp_linear	40.04	30.85	28.60	35.38	32.98	27.62	32.58
elst	42.24	30.01	28.92	34.99	31.43	26.53	32.35
dt	42.18	31.61	33.15	37.65	34.14	28.61	34.56
dkl	45.06	28.40	31.41	33.16	29.94	27.75	32.62
gp_rbf_gpytorch	41.58	24.03	29.14	34.49	27.48	22.02	29.79
gp_sm_gpytorch	43.55	29.69	31.38	38.27	30.87	25.95	33.28
gp_rbf_gpy	43.32	28.10	30.97	36.65	29.39	23.26	31.95
gp_rbf_torch	53.94	31.26	35.00	37.57	31.79	24.73	35.72
gp_rbf	43.18	23.67	27.98	33.41	26.83	21.69	29.46
nsgp_rbf	50.15	29.27	31.13	35.92	32.61	25.89	34.16
mlp	42.54	30.23	30.43	36.61	31.20	27.41	33.07
mlp_gen	40.00	--	--	--	--	--	--

R^2 Score

	fold_0	fold_1	fold_2	fold_3	fold_4	fold_5	average
rf	0.79	0.92	0.87	0.84	0.88	0.91	0.87
svr	0.75	0.91	0.88	0.84	0.88	0.89	0.86
gp_m32	0.76	0.90	0.86	0.59	0.89	0.91	0.82
gp_m12	0.77	0.91	0.87	0.82	0.88	0.91	0.86
gp_linear	0.78	0.88	0.88	0.82	0.85	0.87	0.85
elst	0.76	0.88	0.87	0.83	0.86	0.88	0.85
dt	0.76	0.87	0.83	0.80	0.84	0.86	0.83
dkl	0.73	0.90	0.85	0.84	0.88	0.87	0.84
gp_rbf_gpytorch	0.77	0.93	0.87	0.83	0.90	0.92	0.87
gp_sm_gpytorch	0.75	0.89	0.85	0.79	0.87	0.89	0.84
mlp	0.76	0.88	0.86	0.81	0.87	0.87	0.84
mlp_gen	0.78	--	--	--	--	--	--