-
Notifications
You must be signed in to change notification settings - Fork 1
/
varModel.py
289 lines (235 loc) · 11.1 KB
/
varModel.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
from statsmodels.tsa.stattools import adfuller
import pandas as pd
from statsmodels.tsa.api import VAR
import matplotlib.pyplot as plt
from statsmodels.tsa.arima.model import ARIMA
from math import sqrt
from sklearn.metrics import mean_squared_error
PATH_1min = '../Data/preFinal_1min_v2.csv'
PATH_5min = '../Data/preFinal_5min_v2.csv'
PATH_20min = '../Data/preFinal_20min_v2.csv'
PATH_30min = '../Data/preFinal_30min_v2.csv'
PATH_1h = '../Data/preFinal_1h_v2.csv'
def importData(PATH_1min, PATH_5min, PATH_20min, PATH_30min, PATH_1h):
print("Reading in data...")
df_1min = pd.read_csv(PATH_1min)
df_5min = pd.read_csv(PATH_5min)
df_20min = pd.read_csv(PATH_20min)
df_30min = pd.read_csv(PATH_30min)
df_1h = pd.read_csv(PATH_1h)
print("Data extracted!")
return df_1min, df_5min, df_20min, df_30min, df_1h
def interpol(df_1min, df_5min, df_20min, df_30min, df_1h):
print("Interpolating")
# Interpolating df_1min
df_1min['speed'].interpolate(inplace=True)
df_1min['pressure'].interpolate(method="akima", inplace=True)
df_1min['pressure'].fillna(method='bfill', inplace=True)
df_1min['rel_angle'].interpolate(method="akima", inplace=True)
df_1min['temperature'].interpolate(method="akima", inplace=True)
# Interpolating df_5min
df_5min['speed'].interpolate(inplace=True)
df_5min['pressure'].interpolate(method="akima", inplace=True)
df_5min['pressure'].fillna(method='bfill', inplace=True)
df_5min['rel_angle'].interpolate(method="akima", inplace=True)
df_5min['temperature'].interpolate(method="akima", inplace=True)
# Interpolating df_20min
df_20min['speed'].interpolate(inplace=True)
df_20min['pressure'].interpolate(method='akima', inplace=True)
df_20min['pressure'].fillna(method='bfill', inplace=True)
df_20min['rel_angle'].interpolate(method="akima", inplace=True)
df_20min['temperature'].interpolate(method="akima", inplace=True)
# Interpolating df_30min
df_30min['speed'].interpolate(inplace=True)
df_30min['pressure'].interpolate(method="akima", inplace=True)
df_30min['pressure'].fillna(method='bfill', inplace=True)
df_30min['rel_angle'].interpolate(method="akima", inplace=True)
df_30min['temperature'].interpolate(method="akima", inplace=True)
# Interpolating df_1h
df_1h['speed'].interpolate(inplace=True)
df_1h['pressure'].interpolate(method="akima", inplace=True)
df_1h['pressure'].fillna(method='bfill', inplace=True)
df_1h['rel_angle'].interpolate(method="akima", inplace=True)
df_1h['temperature'].interpolate(method="akima", inplace=True)
print("Interpolated successfully!")
return df_1min, df_5min, df_20min, df_30min, df_1h
def performADF(df):
dftest = adfuller(df, autolag='AIC')
adf = pd.Series(dftest[0:4], index=['Test Statistic', 'p-value', '# of Lags', '# of Observations'])
for key, value in dftest[4].items():
adf['Critical Value (%s)'%key] = value
print(adf)
p = adf['p-value']
if p <= 0.05:
print("\nSeries is Stationary")
else:
print("\nSeries is Non-Stationary")
def differ(ds_train):
ds_train_diff = ds_train.diff().dropna()
return ds_train_diff
def modelVAR(ds_train_diff, n_obs, ds):
model = VAR(ds_train_diff)
results = model.fit(maxlags=15, ic='aic')
print(results.summary())
lag_order = results.k_ar
predicted = results.forecast(ds_train_diff.values[int(-lag_order):], (len(ds.index) - int(n_obs)))
del ds['time']
forecast = pd.DataFrame(predicted, index=ds.index[int(n_obs):], columns=ds.columns)
forecast['pressure'] = (ds['pressure'].iloc[-1] - ds['pressure'].iloc[-2]) + forecast['pressure'].cumsum()
forecast['rel_angle'] = (ds['pressure'].iloc[-1] - ds['pressure'].iloc[-2]) + forecast['pressure'].cumsum()
forecast['temperature'] = (ds['temperature'].iloc[-1] - ds['temperature'].iloc[-2]) + forecast['temperature'].cumsum()
return forecast
def plotVAR(ds, forecast):
fig, axs = plt.subplots(2)
axs[0].plot(ds)
axs[1].plot(forecast)
plt.show()
def main():
# Reading in data
df_1min, df_5min, df_20min, df_30min, df_1h = importData(PATH_1min, PATH_5min, PATH_20min, PATH_30min, PATH_1h)
# Interpolating missing values
df_1min_f, df_5min_f, df_20min_f, df_30min_f, df_1h_f = interpol(df_1min, df_5min, df_20min, df_30min, df_1h)
# Dropping time and timestamp columns as they do not matter. Evaluating observations.
del df_1min_f['timestamp']
del df_1min_f['Unnamed: 0']
n_obs_1min = (len(df_1min_f.index))*0.7
print("1min | N. of Obs.: " + str(n_obs_1min))
del df_5min_f['timestamp']
del df_5min_f['Unnamed: 0']
n_obs_5min = (len(df_5min_f.index))*0.7
print("5min | N. of Obs.: " + str(n_obs_5min))
del df_20min_f['timestamp']
n_obs_20min = (len(df_20min_f.index))*0.7
print("20min | N. of Obs.: " + str(n_obs_20min))
del df_30min_f['timestamp']
n_obs_30min = (len(df_30min_f.index))*0.7
print("30min | N. of Obs.: " + str(n_obs_30min))
del df_1h_f['timestamp']
n_obs_1h = (len(df_1h_f.index))*0.7
print("1h | N. of Obs.: " + str(n_obs_1h))
# Split data
ds_1min_train, ds_1min_test = df_1min_f[:int(-n_obs_1min)], df_1min_f[int(-n_obs_1min):]
ds_5min_train, ds_5min_test = df_5min_f[:int(-n_obs_5min)], df_5min_f[int(-n_obs_5min):]
ds_20min_train, ds_20min_test = df_20min_f[:int(-n_obs_20min)], df_20min_f[int(-n_obs_20min):]
ds_30min_train, ds_30min_test = df_30min_f[:int(-n_obs_30min)], df_30min_f[int(-n_obs_30min):]
ds_1h_train, ds_1h_test = df_1h_f[:int(-n_obs_1h)], df_1h_f[int(-n_obs_1h):]
ds_1min_time = ds_1min_train['time']
del ds_1min_train['time']
ds_5min_time = ds_5min_train['time']
del ds_5min_train['time']
ds_20min_time = ds_20min_train['time']
del ds_20min_train['time']
ds_30min_time = ds_30min_train['time']
del ds_30min_train['time']
ds_1h_time = ds_1h_train['time']
del ds_1h_train['time']
# Differencing variables, which are Non-Stationary
speed_1min = pd.DataFrame(ds_1min_train['speed'])
ds_1min_train_diff = differ(ds_1min_train)
ds_1min_train_diff['speed'] = speed_1min
#print(ds_1min_train_diff.head())
speed_5min = pd.DataFrame(ds_5min_train['speed'])
ds_5min_train_diff = differ(ds_5min_train)
ds_5min_train_diff['speed'] = speed_5min
#print(ds_5min_train_diff.head())
speed_20min = pd.DataFrame(ds_20min_train['speed'])
ds_20min_train_diff = differ(ds_20min_train)
ds_20min_train_diff['speed'] = speed_20min
#print(ds_20min_train_diff.head())
speed_30min = pd.DataFrame(ds_30min_train['speed'])
ds_30min_train_diff = differ(ds_30min_train)
ds_30min_train_diff['speed'] = speed_30min
#print(ds_30min_train_diff.head())
speed_1h = pd.DataFrame(ds_1h_train['speed'])
ds_1h_train_diff = differ(ds_1h_train)
ds_1h_train_diff['speed'] = speed_1h
#print(ds_1h_train_diff.head())
# Perform ADF Test
for i in ds_1min_train_diff.columns:
print("1min | Column: ", i)
print("------------------------------------")
#performADF(ds_1min_train_diff[i])
print('\n')
for i in ds_5min_train_diff.columns:
print("5min | Column: ", i)
print("------------------------------------")
#performADF(ds_1min_train_diff[i])
print('\n')
for i in ds_20min_train_diff.columns:
print("20min | Column: ", i)
print("------------------------------------")
#performADF(ds_1min_train_diff[i])
print('\n')
for i in ds_30min_train_diff.columns:
print("30min | Column: ", i)
print("------------------------------------")
#performADF(ds_1min_train_diff[i])
print('\n')
for i in ds_1h_train_diff.columns:
print("1h | Column: ", i)
print("------------------------------------")
#performADF(ds_1min_train_diff[i])
print('\n')
# Reintegrating time
ds_1min_train_diff['time'] = ds_1min_time
ds_1min_train_diff['time'] = pd.to_datetime(ds_1min_train_diff['time'])
#ds_1min_train_diff['time'] = pd.to_numeric(ds_1min_train_diff['time'])
ds_1min_train_diff.index = pd.DatetimeIndex(ds_1min_train_diff['time']).to_period('min')
del ds_1min_train_diff['time']
print(ds_1min_train_diff.head())
ds_5min_train_diff['time'] = ds_5min_time
ds_5min_train_diff['time'] = pd.to_datetime(ds_5min_train_diff['time'])
#ds_5min_train_diff['time'] = pd.to_numeric(ds_5min_train_diff['time'])
ds_5min_train_diff.index = pd.DatetimeIndex(ds_5min_train_diff['time']).to_period('min')
del ds_5min_train_diff['time']
print(ds_5min_train_diff.head())
ds_20min_train_diff['time'] = ds_20min_time
ds_20min_train_diff['time'] = pd.to_datetime(ds_20min_train_diff['time'])
#ds_20min_train_diff['time'] = pd.to_numeric(ds_20min_train_diff['time'])
ds_20min_train_diff.index = pd.DatetimeIndex(ds_20min_train_diff['time']).to_period('min')
del ds_20min_train_diff['time']
print(ds_20min_train_diff.head())
ds_30min_train_diff['time'] = ds_30min_time
ds_30min_train_diff['time'] = pd.to_datetime(ds_30min_train_diff['time'])
#ds_30min_train_diff['time'] = pd.to_numeric(ds_30min_train_diff['time'])
ds_30min_train_diff.index = pd.DatetimeIndex(ds_30min_train_diff['time']).to_period('min')
del ds_30min_train_diff['time']
print(ds_30min_train_diff.head())
ds_1h_train_diff['time'] = ds_1h_time
ds_1h_train_diff['time'] = pd.to_datetime(ds_1h_train_diff['time'])
#ds_1h_train_diff['time'] = pd.to_numeric(ds_1h_train_diff['time'])
ds_1h_train_diff.index = pd.DatetimeIndex(ds_1h_train_diff['time']).to_period('min')
del ds_1h_train_diff['time']
print(ds_1h_train_diff.head())
# Setting time as index in test datasets
ds_30min_test.index = pd.DatetimeIndex(ds_30min_test['time']).to_period('min')
# Fitting VAR-Model + Forecasting size of test-set
#forecast = modelVAR(ds_1min_train_diff, n_obs_1min, df_1min_f)
#print(forecast.head())
#plotVAR(df_1min_f, forecast)
# Fitting ARIMA-Model + Forecasting
model = ARIMA(ds_30min_train_diff['speed'], order=(15, 0, 0))
model_fit = model.fit()
print(model_fit.summary())
# Plotting residuals
residuals = pd.DataFrame(model_fit.resid)
residuals.plot()
plt.show()
# density plot of residuals
residuals.plot(kind='kde')
plt.show()
# summary stats of residuals
print(residuals.describe())
output = model_fit.forecast(steps=5)
print(output)
df_output = pd.DataFrame.from_dict(output)
#df_output['predicted_mean'] = float(df_output['predicted_mean'])
#print(df_output['predicted_mean'].dtype)
print(df_output)
print(ds_30min_test.head())
fig, axs = plt.subplots(2)
axs[0].plot(df_output['predicted_mean'])
axs[1].plot(ds_30min_test['speed'].head())
plt.show()
if __name__ == "__main__":
main()