Simulation Tables

ITSTGCN

Author

SEOYEON CHOI

Published

April 27, 2023

Simulation Tables

import

import pandas as pd

data_fivenodes= pd.read_csv('./simulation_results/Real_simulation/fivedones_Simulation.csv')

data_fivenodes['dataset']='fivenodes'

data_chickenpox = pd.read_csv('./simulation_results/Real_simulation/chikenpox_Simulation.csv')

data_chickenpox['dataset'] = 'chickenpox'

data_pedalme = pd.read_csv('./simulation_results/Real_simulation/pedalme_Simulation.csv')

data_wikimath = pd.read_csv('./simulation_results/Real_simulation/wikimath.csv')

data_wikimath_block = pd.read_csv('./simulation_results/Real_simulation/wikimath_block.csv')

data_windmillmedium = pd.read_csv('./simulation_results/Real_simulation/windmillmedium.csv')

data_windmillmedium['dataset']='windmillmedium'

data_windmillsmall = pd.read_csv('./simulation_results/Real_simulation/windmillsmall.csv')

data_windmillsmall['dataset'] = 'windmillsmall'

data_windmillsmall_block = pd.read_csv('./simulation_results/Real_simulation/windmillsmall_block.csv')

data_windmillsmall_block['dataset'] = 'windmillsmall'

data_monte = pd.read_csv('./simulation_results/Real_simulation/monte.csv')

data_monte_block = pd.read_csv('./simulation_results/Real_simulation/monte_block.csv')

data = pd.concat([data_fivenodes,
                 data_chickenpox,
                 data_pedalme,
                 data_wikimath,
                  data_wikimath_block,
                data_windmillmedium,
                 data_windmillsmall,
                  data_windmillsmall_block,
                  data_monte,
                  data_monte_block
                 ],axis=0)

data['RecurrentGCN'] = 'GConvGRU'

data2 = data[['dataset','method','RecurrentGCN','mrate','mtype','lags','nof_filters','inter_method','epoch','mse']]

data2

	dataset	method	RecurrentGCN	mrate	mtype	lags	nof_filters	inter_method	epoch	mse
0	fivenodes	GNAR	GConvGRU	0.700000	rand	2	NaN	linear	NaN	1.406830
1	fivenodes	GNAR	GConvGRU	0.750000	rand	2	NaN	linear	NaN	1.406830
2	fivenodes	GNAR	GConvGRU	0.800000	rand	2	NaN	linear	NaN	1.406830
3	fivenodes	GNAR	GConvGRU	0.850000	rand	2	NaN	linear	NaN	1.406830
4	fivenodes	GNAR	GConvGRU	0.700000	rand	2	NaN	linear	NaN	1.406830
...	...	...	...	...	...	...	...	...	...	...
361	monte	GNAR	GConvGRU	0.149142	block	8	NaN	linear	NaN	1.068464
362	monte	GNAR	GConvGRU	0.149142	block	4	NaN	linear	NaN	1.061937
363	monte	GNAR	GConvGRU	0.149142	block	8	NaN	linear	NaN	1.068464
364	monte	GNAR	GConvGRU	0.149142	block	4	NaN	linear	NaN	1.061937
365	monte	GNAR	GConvGRU	0.149142	block	8	NaN	linear	NaN	1.068464

8283 rows × 10 columns

data2.to_csv('./simulation_results/Real_simulation/allresults.csv',index=False)

data = pd.read_csv('./simulation_results/Real_simulation/allresults.csv')

data_DCRNN_fivenodes = pd.read_csv('./simulation_results/DCRNN/fivenodes_DVRNN.csv')

data_DCRNN_chickenpox = pd.read_csv('./simulation_results/DCRNN/chickenpox_DCRNN.csv')

data_DCRNN_pedalme = pd.read_csv('./simulation_results/DCRNN/pedalme_DVRNN.csv')

data_DCRNN_wikimath = pd.read_csv('./simulation_results/DCRNN/wikimath_DCRNN.csv')

data_DCRNN_windmillsmall = pd.read_csv('./simulation_results/DCRNN/windmillsmall_DCRNN.csv')

Fivenodes

Baseline

data.query("dataset=='fivenodes' and mtype!='block'")['mrate'].unique()

array([0.7 , 0.75, 0.8 , 0.85, 0.  ])

pd.merge(data.query("dataset=='fivenodes' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','method','lags'])['mse'].mean().reset_index(),
         data.query("dataset=='fivenodes' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','method','lags'])['mse'].std().reset_index(),
         on=['method','nof_filters','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	nof_filters	method	lags	mean	std
0	12.0	IT-STGCN	2	1.168	0.030
1	12.0	STGCN	2	1.173	0.036
2	16.0	IT-STGCN	2	1.166	0.039
3	16.0	STGCN	2	1.165	0.040

pd.merge(data.query("dataset=='fivenodes' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','lags'])['mse'].mean().reset_index(),
         data.query("dataset=='fivenodes' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','lags'])['mse'].std().reset_index(),
         on=['nof_filters','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	nof_filters	lags	mean	std
0	12.0	2	1.170	0.033
1	16.0	2	1.165	0.039

pd.merge(data_DCRNN_fivenodes.query("dataset=='fivenodes' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','lags'])['mse'].mean().reset_index(),
         data_DCRNN_fivenodes.query("dataset=='fivenodes' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','lags'])['mse'].std().reset_index(),
         on=['nof_filters','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	nof_filters	lags	mean	std
0	16	2	1.247	0.005

Random

data.query("dataset=='fivenodes' and mtype=='rand'and method=='GNAR'")['mse'].unique().round(3)

array([1.407])

pd.merge(data.query("dataset=='fivenodes' and mtype=='rand'").groupby(['mrate','nof_filters','method','lags'])['mse'].mean().reset_index(),
         data.query("dataset=='fivenodes' and mtype=='rand'").groupby(['mrate','nof_filters','method','lags'])['mse'].std().reset_index(),
         on=['method','nof_filters','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("nof_filters==12")

	mrate	nof_filters	method	lags	mean	std
0	0.70	12.0	IT-STGCN	2	1.200	0.070
1	0.70	12.0	STGCN	2	1.213	0.083
4	0.75	12.0	IT-STGCN	2	1.188	0.060
5	0.75	12.0	STGCN	2	1.239	0.102
8	0.80	12.0	IT-STGCN	2	1.221	0.083
9	0.80	12.0	STGCN	2	1.226	0.105
12	0.85	12.0	IT-STGCN	2	1.227	0.085
13	0.85	12.0	STGCN	2	1.291	0.252

pd.merge(data.query("dataset=='fivenodes' and mtype=='rand'").groupby(['mrate','nof_filters','method','lags'])['mse'].mean().reset_index(),
         data.query("dataset=='fivenodes' and mtype=='rand'").groupby(['mrate','nof_filters','method','lags'])['mse'].std().reset_index(),
         on=['method','nof_filters','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("nof_filters!=12")

	mrate	nof_filters	method	lags	mean	std
2	0.70	16.0	IT-STGCN	2	1.201	0.068
3	0.70	16.0	STGCN	2	1.227	0.094
6	0.75	16.0	IT-STGCN	2	1.231	0.110
7	0.75	16.0	STGCN	2	1.201	0.072
10	0.80	16.0	IT-STGCN	2	1.232	0.092
11	0.80	16.0	STGCN	2	1.292	0.148
14	0.85	16.0	IT-STGCN	2	1.286	0.297
15	0.85	16.0	STGCN	2	1.362	0.239

pd.merge(data_DCRNN_fivenodes.query("dataset=='fivenodes' and mtype=='rand'").groupby(['mrate','nof_filters','method','lags'])['mse'].mean().reset_index(),
         data_DCRNN_fivenodes.query("dataset=='fivenodes' and mtype=='rand'").groupby(['mrate','nof_filters','method','lags'])['mse'].std().reset_index(),
         on=['method','nof_filters','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("nof_filters!=12 and mrate!=0.3")

	mrate	nof_filters	method	lags	mean	std
2	0.8	16	IT-STGCN	2	1.478	1.245
3	0.8	16	STGCN	2	1.491	0.302

Block

data.query("dataset=='fivenodes' and mtype=='block'and method=='GNAR'")['mse'].unique().round(3)

array([1.407])

data.query("dataset=='fivenodes' and mtype=='block'")

	dataset	method	RecurrentGCN	mrate	mtype	lags	nof_filters	inter_method	epoch	mse
600	fivenodes	GNAR	GConvGRU	0.125	block	2	NaN	cubic	NaN	1.406830
601	fivenodes	GNAR	GConvGRU	0.125	block	2	NaN	linear	NaN	1.406830
602	fivenodes	GNAR	GConvGRU	0.125	block	2	NaN	cubic	NaN	1.406830
603	fivenodes	GNAR	GConvGRU	0.125	block	2	NaN	linear	NaN	1.406830
604	fivenodes	GNAR	GConvGRU	0.125	block	2	NaN	cubic	NaN	1.406830
...	...	...	...	...	...	...	...	...	...	...
967	fivenodes	IT-STGCN	GConvGRU	0.300	block	2	16.0	linear	150.0	1.135442
968	fivenodes	STGCN	GConvGRU	0.300	block	2	12.0	linear	150.0	1.203593
969	fivenodes	STGCN	GConvGRU	0.300	block	2	16.0	linear	150.0	1.220799
970	fivenodes	IT-STGCN	GConvGRU	0.300	block	2	12.0	linear	150.0	1.111655
971	fivenodes	IT-STGCN	GConvGRU	0.300	block	2	16.0	linear	150.0	1.197438

372 rows × 10 columns

pd.merge(data.query("dataset=='fivenodes' and mtype=='block'").groupby(['mrate','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='fivenodes' and mtype=='block'").groupby(['mrate','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','nof_filters','mrate']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	mrate	nof_filters	method	mean	std
0	0.125	12.0	IT-STGCN	4.308	3.333
1	0.125	12.0	STGCN	6.722	5.755
2	0.125	16.0	IT-STGCN	4.633	3.737
3	0.125	16.0	STGCN	6.858	5.814
4	0.300	12.0	IT-STGCN	1.178	0.032
5	0.300	12.0	STGCN	1.232	0.040
6	0.300	16.0	IT-STGCN	1.163	0.050
7	0.300	16.0	STGCN	1.232	0.053

epoch 별 보기

df1 = pd.read_csv('./simulation_results/fivenodes/fivenodes_STGCN_ITSTGCN_random_epoch50.csv')
df2 = pd.read_csv('./simulation_results/fivenodes/fivenodes_STGCN_ITSTGCN_random_epoch100.csv')
df3 = pd.read_csv('./simulation_results/fivenodes/fivenodes_STGCN_ITSTGCN_random_epoch150.csv')
df4 = pd.read_csv('./simulation_results/fivenodes/fivenodes_STGCN_ITSTGCN_random_epoch200.csv')

df_gnar = pd.read_csv('./simulation_results/fivenodes/fivenodes_GNAR_random.csv')

data_temp = pd.concat([df1,df2,df3,df4,df_gnar],axis=0)

STGCN은 nearest에서 mse가 낮았다.

data_temp.query("method=='STGCN' and mtype=='rand' and mrate==0.8 and lags==2 and inter_method=='linear' and nof_filters==4").\
groupby(['method','epoch','mrate','lags','nof_filters','inter_method'])['mse'].mean().reset_index()['mse'].mean()

1.182556539773941

data_temp.query("method=='STGCN' and mtype=='rand' and mrate==0.8 and lags==2 and inter_method=='linear' and nof_filters==4").\
groupby(['method','epoch','mrate','lags','nof_filters','inter_method'])['mse'].mean().reset_index()['mse'].std()

0.012169932740213692

data_temp.query("method=='IT-STGCN' and mtype=='rand' and mrate==0.8 and lags==2 and inter_method=='linear' and nof_filters==4").\
groupby(['method','epoch','mrate','lags','nof_filters','inter_method'])['mse'].mean().reset_index()['mse'].mean()

1.1747438261906304

data_temp.query("method=='IT-STGCN' and mtype=='rand' and mrate==0.8 and lags==2 and inter_method=='linear' and nof_filters==4").\
groupby(['method','epoch','mrate','lags','nof_filters','inter_method'])['mse'].mean().reset_index()['mse'].std()

0.007602895892378366

ChickenpoxDatasetLoader(lags=4)

Baseline

pd.merge(data.query("dataset=='chickenpox' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','method'])['mse'].std().reset_index(),
         on=['method','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("nof_filters==16")

	nof_filters	method	mean	std
0	16.0	IT-STGCN	1.008	0.010
1	16.0	STGCN	1.009	0.008

pd.merge(data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype!='rand' and mtype!='block'").groupby(['nof_filters','method'])['mse'].std().reset_index(),
         on=['method','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("nof_filters==16")

	nof_filters	method	mean	std
0	16	IT-STGCN	0.953	0.005
1	16	STGCN	0.953	0.006

Random

data.query("dataset=='chickenpox' and mtype=='rand'and method=='GNAR'")['mse'].unique().round(3)

array([1.427])

pd.merge(data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.3 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
0	0.3	cubic	16.0	IT-STGCN	1.019	0.011
1	0.3	cubic	16.0	STGCN	1.059	0.013
6	0.3	linear	16.0	IT-STGCN	1.015	0.009
7	0.3	linear	16.0	STGCN	1.040	0.014

pd.merge(data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.3 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
0	0.3	cubic	16	IT-STGCN	0.985	0.008
1	0.3	cubic	16	STGCN	1.053	0.008
2	0.3	linear	16	IT-STGCN	0.983	0.007
3	0.3	linear	16	STGCN	1.028	0.012

pd.merge(data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.4 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
4	0.4	cubic	16	IT-STGCN	0.995	0.009
5	0.4	cubic	16	STGCN	1.069	0.011
6	0.4	linear	16	IT-STGCN	0.994	0.008
7	0.4	linear	16	STGCN	1.038	0.011

pd.merge(data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.4 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
12	0.4	cubic	16.0	IT-STGCN	1.021	0.009
13	0.4	cubic	16.0	STGCN	1.084	0.025
18	0.4	linear	16.0	IT-STGCN	1.020	0.009
19	0.4	linear	16.0	STGCN	1.051	0.014

pd.merge(data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.5 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
24	0.5	cubic	16.0	IT-STGCN	1.027	0.012
25	0.5	cubic	16.0	STGCN	1.128	0.042
30	0.5	linear	16.0	IT-STGCN	1.026	0.014
31	0.5	linear	16.0	STGCN	1.071	0.016

pd.merge(data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.5 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
8	0.5	cubic	16	IT-STGCN	1.011	0.007
9	0.5	cubic	16	STGCN	1.080	0.019
10	0.5	linear	16	IT-STGCN	1.008	0.007
11	0.5	linear	16	STGCN	1.055	0.010

pd.merge(data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.8 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
36	0.8	cubic	16.0	IT-STGCN	1.206	0.117
37	0.8	cubic	16.0	STGCN	1.266	0.152
42	0.8	linear	16.0	IT-STGCN	1.101	0.034
43	0.8	linear	16.0	STGCN	1.166	0.059

pd.merge(data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.8 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
12	0.8	cubic	16	IT-STGCN	1.181	0.142
13	0.8	cubic	16	STGCN	1.417	0.663
14	0.8	linear	16	IT-STGCN	1.058	0.015
15	0.8	linear	16	STGCN	1.102	0.027

pd.merge(data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.9 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
48	0.9	cubic	16.0	IT-STGCN	1.228	0.199
49	0.9	cubic	16.0	STGCN	1.283	0.222
54	0.9	linear	16.0	IT-STGCN	1.251	0.106
55	0.9	linear	16.0	STGCN	1.265	0.148

pd.merge(data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_chickenpox.query("dataset=='chickenpox' and mtype=='rand'").groupby(['mrate','inter_method','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.9 and nof_filters==16")

	mrate	inter_method	nof_filters	method	mean	std
16	0.9	cubic	16	IT-STGCN	2.372	1.841
17	0.9	cubic	16	STGCN	1.596	0.648
18	0.9	linear	16	IT-STGCN	1.090	0.045
19	0.9	linear	16	STGCN	1.179	0.127

Block

data.query("dataset=='chickenpox' and mtype=='block'and method=='GNAR'")['mse'].unique()

array([1.42749429])

pd.merge(data.query("dataset=='chickenpox' and mtype=='block'").groupby(['inter_method','mrate','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='chickenpox' and mtype=='block'").groupby(['inter_method','mrate','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','inter_method','mrate','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("nof_filters==16")

	inter_method	mrate	nof_filters	method	mean	std
0	cubic	0.288	16.0	IT-STGCN	1.052	0.028
1	cubic	0.288	16.0	STGCN	1.052	0.023
6	linear	0.288	16.0	IT-STGCN	1.008	0.005
7	linear	0.288	16.0	STGCN	1.011	0.008

PedalMeDatasetLoader (lags=4)

Baseline

pd.merge(data.query("dataset=='pedalme' and mtype!='rand' and mtype!='block'").groupby(['lags','nof_filters','method'])['mse'].mean().reset_index(),
         data.query("dataset=='pedalme' and mtype!='rand' and mtype!='block'").groupby(['lags','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','lags','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("lags==4")

	lags	nof_filters	method	mean	std
0	4	12.0	IT-STGCN	1.241	0.04
1	4	12.0	STGCN	1.271	0.04

(1.241+1.271)/2

1.256

pd.merge(data_DCRNN_pedalme.query("dataset=='pedalme' and mtype!='rand' and mtype!='block'").groupby(['lags','nof_filters','method'])['mse'].mean().reset_index(),
         data_DCRNN_pedalme.query("dataset=='pedalme' and mtype!='rand' and mtype!='block'").groupby(['lags','nof_filters','method'])['mse'].std().reset_index(),
         on=['method','lags','nof_filters']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	lags	nof_filters	method	mean	std
0	4	12	IT-STGCN	1.204	0.020
1	4	12	STGCN	1.203	0.022

(1.204+1.203)/2

1.2035

Random

data.query("dataset=='pedalme' and method=='GNAR' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index().query(" lags==4")

	mrate	lags	inter_method	method	mse
0	0.3	4	cubic	GNAR	1.302679
1	0.3	4	linear	GNAR	1.302679
2	0.4	4	cubic	GNAR	1.302679
3	0.4	4	linear	GNAR	1.302679
4	0.5	4	cubic	GNAR	1.302679
5	0.5	4	linear	GNAR	1.302679
6	0.6	4	cubic	GNAR	1.302679
7	0.6	4	linear	GNAR	1.302679

pd.merge(data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.3 and lags==4")

	mrate	lags	inter_method	method	mean	std
0	0.3	4	cubic	GNAR	1.303	0.000
1	0.3	4	cubic	IT-STGCN	1.314	0.109
2	0.3	4	cubic	STGCN	1.363	0.115
3	0.3	4	linear	GNAR	1.303	0.000
4	0.3	4	linear	IT-STGCN	1.323	0.094
5	0.3	4	linear	STGCN	1.380	0.127

pd.merge(data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.3 and  lags==4")

	mrate	lags	inter_method	method	mean	std
0	0.3	4	cubic	IT-STGCN	1.223	0.031
1	0.3	4	cubic	STGCN	1.248	0.039
2	0.3	4	linear	IT-STGCN	1.227	0.026
3	0.3	4	linear	STGCN	1.242	0.031
4	0.3	4	nearest	IT-STGCN	1.231	0.032
5	0.3	4	nearest	STGCN	1.229	0.032

pd.merge(data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.4 and lags==4")

	mrate	lags	inter_method	method	mean	std
10	0.4	4	cubic	GNAR	1.303	0.000
11	0.4	4	cubic	IT-STGCN	1.331	0.112
12	0.4	4	cubic	STGCN	1.342	0.108
13	0.4	4	linear	GNAR	1.303	0.000
14	0.4	4	linear	IT-STGCN	1.375	0.154
15	0.4	4	linear	STGCN	1.397	0.193

pd.merge(data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query(" mrate==0.4 and lags==4")

	mrate	lags	inter_method	method	mean	std
6	0.4	4	cubic	IT-STGCN	1.230	0.030
7	0.4	4	cubic	STGCN	1.257	0.051
8	0.4	4	linear	IT-STGCN	1.231	0.032
9	0.4	4	linear	STGCN	1.251	0.040
10	0.4	4	nearest	IT-STGCN	1.235	0.031
11	0.4	4	nearest	STGCN	1.241	0.033

pd.merge(data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.5 and lags==4")

	mrate	lags	inter_method	method	mean	std
20	0.5	4	cubic	GNAR	1.303	0.000
21	0.5	4	cubic	IT-STGCN	1.328	0.108
22	0.5	4	cubic	STGCN	1.367	0.114
23	0.5	4	linear	GNAR	1.303	0.000
24	0.5	4	linear	IT-STGCN	1.377	0.138
25	0.5	4	linear	STGCN	1.326	0.129

pd.merge(data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query(" mrate==0.5 and lags==4")

	mrate	lags	inter_method	method	mean	std
12	0.5	4	cubic	IT-STGCN	1.251	0.034
13	0.5	4	cubic	STGCN	1.279	0.095
14	0.5	4	linear	IT-STGCN	1.241	0.037
15	0.5	4	linear	STGCN	1.268	0.052
16	0.5	4	nearest	IT-STGCN	1.245	0.034
17	0.5	4	nearest	STGCN	1.256	0.043

pd.merge(data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.6 and lags==4")

	mrate	lags	inter_method	method	mean	std
30	0.6	4	cubic	GNAR	1.303	0.000
31	0.6	4	cubic	IT-STGCN	1.300	0.063
32	0.6	4	cubic	STGCN	1.352	0.106
33	0.6	4	linear	GNAR	1.303	0.000
34	0.6	4	linear	IT-STGCN	1.416	0.169
35	0.6	4	linear	STGCN	1.326	0.106

pd.merge(data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data_DCRNN_pedalme.query("dataset=='pedalme' and mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("mrate==0.6 and  lags==4")

	mrate	lags	inter_method	method	mean	std
18	0.6	4	cubic	IT-STGCN	1.259	0.052
19	0.6	4	cubic	STGCN	1.313	0.193
20	0.6	4	linear	IT-STGCN	1.243	0.036
21	0.6	4	linear	STGCN	1.280	0.064
22	0.6	4	nearest	IT-STGCN	1.254	0.037
23	0.6	4	nearest	STGCN	1.271	0.050

Block

pd.merge(data.query("dataset=='pedalme' and mtype=='block'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data.query("dataset=='pedalme' and mtype=='block'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("lags==4")

	mrate	lags	inter_method	method	mean	std
4	0.143	4	cubic	GNAR	1.303	0.000
5	0.143	4	cubic	IT-STGCN	1.284	0.053
6	0.143	4	cubic	STGCN	1.288	0.071
7	0.143	4	linear	GNAR	1.303	0.000
14	0.286	4	cubic	GNAR	1.303	0.000
15	0.286	4	cubic	IT-STGCN	1.304	0.050
16	0.286	4	cubic	STGCN	1.377	0.061
17	0.286	4	linear	GNAR	1.303	0.000
18	0.286	4	linear	IT-STGCN	1.335	0.062
19	0.286	4	linear	STGCN	1.350	0.056

W_st

data_pedalme_wst = pd.read_csv('./simulation_results/Real_simulation/pedalme_Simulation_itstgcnsnd.csv')

pd.merge(data_pedalme_wst.query("mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data_pedalme_wst.query("mtype=='rand'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("lags==4")

	mrate	lags	inter_method	method	mean	std
0	0.3	4	cubic	IT-STGCN	1.353	0.141
1	0.3	4	cubic	STGCN	1.360	0.131
2	0.3	4	linear	IT-STGCN	1.337	0.122
3	0.3	4	linear	STGCN	1.353	0.117
4	0.3	4	nearest	IT-STGCN	1.316	0.122
5	0.3	4	nearest	STGCN	1.403	0.134
12	0.4	4	cubic	IT-STGCN	1.332	0.166
13	0.4	4	cubic	STGCN	1.344	0.123
14	0.4	4	linear	IT-STGCN	1.355	0.139
15	0.4	4	linear	STGCN	1.393	0.168
16	0.4	4	nearest	IT-STGCN	1.386	0.128
17	0.4	4	nearest	STGCN	1.341	0.129
24	0.5	4	cubic	IT-STGCN	1.312	0.152
25	0.5	4	cubic	STGCN	1.362	0.129
26	0.5	4	linear	IT-STGCN	1.344	0.177
27	0.5	4	linear	STGCN	1.335	0.117
28	0.5	4	nearest	IT-STGCN	1.335	0.153
29	0.5	4	nearest	STGCN	1.350	0.129
36	0.6	4	cubic	IT-STGCN	1.346	0.151
37	0.6	4	cubic	STGCN	1.398	0.103
38	0.6	4	linear	IT-STGCN	1.365	0.177
39	0.6	4	linear	STGCN	1.353	0.087
40	0.6	4	nearest	IT-STGCN	1.402	0.269
41	0.6	4	nearest	STGCN	1.339	0.111
48	0.7	4	cubic	IT-STGCN	1.377	0.173
49	0.7	4	cubic	STGCN	1.363	0.097
50	0.7	4	linear	IT-STGCN	1.355	0.144
51	0.7	4	linear	STGCN	1.288	0.063
52	0.7	4	nearest	IT-STGCN	1.383	0.157
53	0.7	4	nearest	STGCN	1.334	0.124

pd.merge(data_pedalme_wst.query("mtype=='block'").groupby(['mrate','lags','inter_method','method'])['mse'].mean().reset_index(),
         data_pedalme_wst.query("mtype=='block'").groupby(['mrate','lags','inter_method','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags','inter_method']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("lags==4")

	mrate	lags	inter_method	method	mean	std
6	0.286	4	cubic	IT-STGCN	1.260	0.063
7	0.286	4	cubic	STGCN	1.417	0.065
8	0.286	4	linear	IT-STGCN	1.276	0.065
9	0.286	4	linear	STGCN	1.288	0.055
10	0.286	4	nearest	IT-STGCN	1.275	0.061
11	0.286	4	nearest	STGCN	1.312	0.061

WikiMathsDatasetLoader (lags=8)

Baseline

data.query("dataset=='wikimath' and mrate==0").groupby(['lags','nof_filters','method'])['mse'].mean().reset_index().round(3).query("lags==8")

	lags	nof_filters	method	mse
4	8	12.0	IT-STGCN	0.771
5	8	12.0	STGCN	0.772

data_DCRNN_wikimath.query("dataset=='wikimath' and mrate==0").groupby(['lags','nof_filters','method'])['mse'].mean().reset_index().round(3).query("lags==8")

	lags	nof_filters	method	mse
0	8	12	IT-STGCN	0.778
1	8	12	STGCN	0.759

Random

pd.merge(data.query("dataset=='wikimath' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='wikimath' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("lags==8")

	mrate	lags	method	mean	std
6	0.3	8	IT-STGCN	0.781	0.012
7	0.3	8	STGCN	0.779	0.013
14	0.5	8	IT-STGCN	0.802	0.041
15	0.5	8	STGCN	0.806	0.020

pd.merge(data_DCRNN_wikimath.query("dataset=='wikimath' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data_DCRNN_wikimath.query("dataset=='wikimath' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3).query("lags==8")

	mrate	lags	method	mean	std
0	0.3	8	IT-STGCN	0.759	0.021
1	0.3	8	STGCN	0.774	0.030

Block

pd.merge(data.query("dataset=='wikimath' and mtype=='block'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='wikimath' and mtype=='block'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'})

	mrate	lags	method	mean	std
0	0.003841	2	IT-STGCN	0.810475	0.033897
1	0.003841	2	STGCN	0.801502	0.015510
2	0.003841	4	IT-STGCN	0.779852	0.013188
3	0.003841	4	STGCN	0.779816	0.019309

missing values on the same nodes

data_wikimath_st = pd.read_csv('./simulation_results/Real_simulation/wikimath_GSO_st.csv')

pd.merge(data_wikimath_st.groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
        data_wikimath_st.groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	mrate	lags	method	mean	std
0	0.123	4	IT-STGCN	0.774	0.008
1	0.123	4	STGCN	0.766	0.010
2	0.738	4	IT-STGCN	0.851	0.029
3	0.738	4	STGCN	0.831	0.031

WindmillOutputSmallDatasetLoader (lags=8)

Baseline

pd.merge(data.query("dataset=='windmillsmall' and mrate==0").groupby(['lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='windmillsmall' and mrate==0").groupby(['lags','method'])['mse'].std().reset_index(),
         on=['method','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	lags	method	mean	std
0	8	GNAR	1.649	0.000
1	8	IT-STGCN	1.006	0.006
2	8	STGCN	1.001	0.003

pd.merge(data_DCRNN_windmillsmall.query("dataset=='windmillsmall' and mrate==0").groupby(['lags','method'])['mse'].mean().reset_index(),
         data_DCRNN_windmillsmall.query("dataset=='windmillsmall' and mrate==0").groupby(['lags','method'])['mse'].std().reset_index(),
         on=['method','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	lags	method	mean	std
0	8	IT-STGCN	1.000	NaN
1	8	STGCN	1.001	NaN

Random

data.query("dataset=='windmillsmall' and mtype=='rand'")

	dataset	method	RecurrentGCN	mrate	mtype	lags	nof_filters	inter_method	epoch	mse
7097	windmillsmall	STGCN	GConvGRU	0.7	rand	8	12.0	linear	50.0	1.436077
7098	windmillsmall	IT-STGCN	GConvGRU	0.7	rand	8	12.0	linear	50.0	1.290896
7099	windmillsmall	STGCN	GConvGRU	0.7	rand	8	12.0	linear	50.0	1.410098
7100	windmillsmall	IT-STGCN	GConvGRU	0.7	rand	8	12.0	linear	50.0	1.176981
7101	windmillsmall	STGCN	GConvGRU	0.7	rand	8	12.0	linear	50.0	1.447851
...	...	...	...	...	...	...	...	...	...	...
7180	windmillsmall	GNAR	GConvGRU	0.7	rand	8	NaN	linear	NaN	1.649230
7181	windmillsmall	GNAR	GConvGRU	0.7	rand	8	NaN	linear	NaN	1.649230
7182	windmillsmall	GNAR	GConvGRU	0.7	rand	8	NaN	linear	NaN	1.649230
7183	windmillsmall	GNAR	GConvGRU	0.7	rand	8	NaN	linear	NaN	1.649230
7184	windmillsmall	GNAR	GConvGRU	0.7	rand	8	NaN	linear	NaN	1.649230

88 rows × 10 columns

pd.merge(data.query("dataset=='windmillsmall' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='windmillsmall' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	mrate	lags	method	mean	std
0	0.7	8	GNAR	1.649	0.000
1	0.7	8	IT-STGCN	1.178	0.054
2	0.7	8	STGCN	1.410	0.075

pd.merge(data_DCRNN_windmillsmall.query("dataset=='windmillsmall' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data_DCRNN_windmillsmall.query("dataset=='windmillsmall' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	mean	mrate	lags	method	std

Block

pd.merge(data.query("dataset=='windmillsmall' and mtype=='block'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='windmillsmall' and mtype=='block'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'}).round(3)

	mrate	lags	method	mean	std
0	0.325	8	GNAR	1.649	0.000
1	0.325	8	IT-STGCN	1.015	0.009
2	0.325	8	STGCN	1.017	0.008

Montevideobus (lags=4)

Baseline

round(data.query("dataset=='monte' and mrate==0")['mse'].mean(),3),round(data_monte.query("mrate==0")['mse'].std(),3)

(0.97, 0.024)

Random

data.query("dataset=='monte' and mtype=='rand'")

	dataset	method	RecurrentGCN	mrate	mtype	lags	nof_filters	inter_method	epoch	mse
7307	monte	STGCN	GConvGRU	0.3	rand	8	12.0	linear	50.0	0.972491
7308	monte	IT-STGCN	GConvGRU	0.3	rand	8	12.0	linear	50.0	0.974410
7309	monte	STGCN	GConvGRU	0.3	rand	8	12.0	linear	50.0	0.968628
7310	monte	IT-STGCN	GConvGRU	0.3	rand	8	12.0	linear	50.0	0.976796
7311	monte	STGCN	GConvGRU	0.3	rand	8	12.0	linear	50.0	0.973314
...	...	...	...	...	...	...	...	...	...	...
7872	monte	IT-STGCN	GConvGRU	0.9	rand	8	12.0	linear	50.0	1.030140
7873	monte	STGCN	GConvGRU	0.9	rand	8	12.0	linear	50.0	1.040731
7874	monte	IT-STGCN	GConvGRU	0.9	rand	8	12.0	linear	50.0	1.041819
7875	monte	STGCN	GConvGRU	0.9	rand	8	12.0	linear	50.0	1.023531
7876	monte	IT-STGCN	GConvGRU	0.9	rand	8	12.0	linear	50.0	1.032515

504 rows × 10 columns

pd.merge(data.query("dataset=='monte' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='monte' and mtype=='rand'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'})

	mrate	lags	method	mean	std
0	0.3	4	GNAR	1.061937	0.000000
1	0.3	4	IT-STGCN	0.971925	0.001871
2	0.3	4	STGCN	0.965885	0.002552
3	0.3	8	GNAR	1.068464	0.000000
4	0.3	8	IT-STGCN	0.974867	0.003233
5	0.3	8	STGCN	0.972051	0.002383
6	0.4	4	GNAR	1.061937	0.000000
7	0.4	4	IT-STGCN	0.976348	0.001374
8	0.4	4	STGCN	0.967480	0.002974
9	0.4	8	GNAR	1.068464	0.000000
10	0.4	8	IT-STGCN	0.978809	0.002110
11	0.4	8	STGCN	0.973098	0.002613
12	0.8	4	GNAR	1.061937	0.000000
13	0.8	4	IT-STGCN	1.006583	0.003297
14	0.8	4	STGCN	0.999715	0.006909
15	0.8	8	GNAR	1.068464	0.000000
16	0.8	8	IT-STGCN	1.004450	0.002953
17	0.8	8	STGCN	1.005238	0.005777
18	0.9	4	GNAR	1.061937	0.000000
19	0.9	4	IT-STGCN	1.034162	0.005611
20	0.9	4	STGCN	1.034995	0.006551
21	0.9	8	GNAR	1.068464	0.000000
22	0.9	8	IT-STGCN	1.030283	0.007979
23	0.9	8	STGCN	1.031538	0.009285

Block

pd.merge(data.query("dataset=='monte' and mtype=='block'").groupby(['mrate','lags','method'])['mse'].mean().reset_index(),
         data.query("dataset=='monte' and mtype=='block'").groupby(['mrate','lags','method'])['mse'].std().reset_index(),
         on=['method','mrate','lags']).rename(columns={'mse_x':'mean','mse_y':'std'})

	mrate	lags	method	mean	std
0	0.149142	4	GNAR	1.061937	0.000000
1	0.149142	4	IT-STGCN	0.963990	0.002194
2	0.149142	4	STGCN	0.965297	0.001611
3	0.149142	8	GNAR	1.068464	0.000000
4	0.149142	8	IT-STGCN	0.971647	0.002860
5	0.149142	8	STGCN	0.971700	0.001672