def test_corner_cases():
assert_raises(ValueError,
auto_arima,
wineind,
error_action='some-bad-string')
# things that produce warnings
with warnings.catch_warnings(record=False):
warnings.simplefilter('ignore')
# show a constant result will result in a quick fit
_ = auto_arima(np.ones(10), suppress_warnings=True)
# show the same thing with return_all results in the ARIMA in a list
_ = auto_arima(np.ones(10),
suppress_warnings=True,
return_valid_fits=True)
assert hasattr(_, '__iter__')
# we did this in 0.1-alpha:
# show that with <= 3 samples, using a non-aic metric reverts to AIC
# try:
# _ = auto_arima(np.arange(3), information_criterion='bic', seasonal=False, suppress_warnings=True)
# except ValueError: # this happens because it can't fit such small data...
# pass
# show we fail for n_iter < 0
assert_raises(ValueError, auto_arima, np.ones(10), random=True, n_fits=-1)
# show if max* < start* it breaks:
assert_raises(ValueError, auto_arima, np.ones(10), start_p=5, max_p=0)
def test_small_samples():
# if n_samples < 10, test the new starting p, d, Q
samp = lynx[:8]
auto_arima(samp,
suppress_warnings=True,
stepwise=True,
error_action='ignore')
def test_with_seasonality6():
# show that we can fit an ARIMA where the max_p|q == start_p|q
auto_arima(hr,
start_p=0,
max_p=0,
d=0,
start_q=0,
max_q=0,
seasonal=False,
max_order=np.inf,
suppress_warnings=True)
def test_failing_model_fit():
with pytest.raises(ValueError):
# raise ValueError('non-invertible starting MA parameters found'
auto_arima(wineind,
seasonal=True,
suppress_warnings=True,
error_action='raise',
m=2,
random=True,
random_state=1,
n_fits=2)
def test_many_orders():
lam = 0.5
lynx_bc = ((lynx**lam) - 1) / lam
auto_arima(lynx_bc,
start_p=1,
start_q=1,
d=0,
max_p=5,
max_q=5,
suppress_warnings=True,
stepwise=True)
def autoarima(self, data, pre_len=7):
D_f = nsdiffs(data, m=3, max_D=5, test='ch')
d_f = ndiffs(data, alpha=0.05, test='kpss', max_d=5)
if len(data) <= 30:
seasonal = False
else:
seasonal = True
try:
stepwise_fit = auto_arima(
data,
start_p=0,
start_q=0,
max_p=3,
max_q=3,
m=12,
start_P=0,
seasonal=seasonal,
d=d_f,
D=D_f,
trace=False,
error_action=
'ignore', # don't want to know if an order does not work
suppress_warnings=True, # don't want convergence warnings
stepwise=True) # set to stepwise
except:
stepwise_fit = auto_arima(
data,
start_p=0,
start_q=0,
max_p=3,
max_q=0,
m=12,
start_P=0,
seasonal=False,
d=0,
D=0,
trace=False,
error_action=
'ignore', # don't want to know if an order does not work
suppress_warnings=True, # don't want convergence warnings
stepwise=True) # set to stepwise
output = stepwise_fit.predict(n_periods=pre_len).tolist()
self._get_model({
'model_name': 'autoarima',
'model': stepwise_fit,
'pred': output,
'org_data': data,
'pre_len': pre_len
})
return output
def train(self, y_train, order=None, seasonal_order=None):
# y_train: training dataset
# order: ARIMA order. Example: (1,1,0)
# seasonal: ARIMA seasonal order. Example: (0,1,0)
stepwise_model = auto_arima(y_train, start_p=0, start_q=1,
max_p=6, max_q=3, m=12,
start_P=0, seasonal=True,
#d=1, D=1,
trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True) #.fit(y_train)
warnings.filterwarnings("ignore") # specify to ignore warning messages
self.model = sm.tsa.statespace.SARIMAX(y_train,
order=stepwise_model.order,
seasonal_order=stepwise_model.seasonal_order,
enforce_stationarity=False,
enforce_invertibility=False
).fit(disp=False)
self.model_params = (stepwise_model.order, stepwise_model.seasonal_order)
return self.model
def gaojier(differsets):
forecast_result = []
for loop in np.arange(len(differsets)):
data = differsets[loop]
fittedmodel = auto_arima(
data,
start_p=1,
start_q=1,
max_p=6,
max_q=6,
max_d=6,
max_order=None,
seasonal=False,
m=1,
test='adf',
trace=False,
error_action=
'ignore', # don't want to know if an order does not work
suppress_warnings=True, # don't want convergence warnings
stepwise=True,
information_criterion='bic',
njob=-1) # set to stepwise
y_hat = fittedmodel.predict(1)[0]
forecast_result.append(y_hat)
return forecast_result
def a_arima(timeseries):
stepwise_model = auto_arima(timeseries,
start_p=1,
start_q=1,
max_p=3,
max_q=3,
m=12,
start_P=0,
seasonal=True,
d=1,
D=1,
trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
print(stepwise_model.aic())
# Train the Model
stepwise_model.fit(train)
# Forecast
future_forecast = stepwise_model.predict(n_periods=182)
future_forecast
future_forecast = pd.DataFrame(future_forecast,
index=test.index,
column=['Prediction'])
pd.concat([test, future_forecast], axis=1).iplot()
def fit(self):
self.model = auto_arima(self.__data, start_p=1, start_q=1,
max_p=3, max_q=3, m=12,
start_P=0, seasonal=True,
d=1, D=1, trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
print('Best found AIC: %f' % (self.model.aic()))
print('Arima model (%d, %d, %d) x (%d, %d, %d, %d) ' % (self.model.order[0],
self.model.order[1], self.model.order[2],
self.model.seasonal_order[0], self.model.seasonal_order[1],
self.model.seasonal_order[2], self.model.seasonal_order[3]))
# self.train = self.__data.loc['1985-01-01':'2016-12-01']
# self.test = self.__data.loc['2015-01-01':]
if self._comparing:
line_to_write = 'AIC : {}, model ({}, {}, {}) x ({}, {}, {}, {})'.format(
self.model.aic(), self.model.order[0],
self.model.order[1], self.model.order[2],
self.model.seasonal_order[0], self.model.seasonal_order[1],
self.model.seasonal_order[2], self.model.seasonal_order[3])
self.save_aic(self.__data.index[0], self.__data.index[-1], line_to_write)
self.results = self.model.fit(self.__data)
def auto_arima_predict(data):
l = len(data)
if l < 4293:
train = data[:round(l * 4 / 5)]
valid = data[round(l * 4 / 5):]
else:
train = data[-4293:l - 973]
valid = data[-973:]
training = train['Close']
validation = valid['Close']
model = auto_arima(training, start_p=1, start_q=1, max_p=3, max_q=3, m=12, start_P=0, seasonal=True, d=1, D=1,
trace=True, error_action='ignore', suppress_warnings=True)
model.fit(training)
forecast = model.predict(n_periods=round(len(data)*1/5))
forecast = pd.DataFrame(forecast, index=valid.index, columns=['Prediction'])
plt.plot(train['Close'])
plt.plot(valid['Close'])
plt.plot(forecast['Prediction'])
plt.show()
print(np.sqrt(np.mean(np.power((np.array(valid['Close'])-np.array(forecast['Prediction'])), 2))))
return forecast['Prediction']
def fit(self, train_size='2016-12-01'):
self.model = auto_arima(self.__data, start_p=1, start_q=1,
max_p=3, max_q=3, m=12,
start_P=0, seasonal=True,
d=1, D=1, trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
print('Best found AIC: %f' % (self.model.aic()))
print('Arima model (%d, %d, %d) x (%d, %d, %d, %d) ' % (self.model.order[0],
self.model.order[1], self.model.order[2],
self.model.seasonal_order[0], self.model.seasonal_order[1],
self.model.seasonal_order[2], self.model.seasonal_order[3]))
# self.train = self.__data.loc['1985-01-01':'2016-12-01']
# self.test = self.__data.loc['2015-01-01':]
self.train = self.__data[:train_size]
if type(train_size) is str:
index = len(self.__data)
index = list(self.__data.index).index(pd.to_datetime(train_size))
if index >= 30:
index -= 30
elif train_size >= 30:
index = train_size - 30
else:
index = train_size
self.test = self.__data[index:]
self.model.fit(self.train)
def auto(self, **kwargs):
"""This method finds the best arima.
@see pyrmid.arima.auto_arima
Parameters
----------
Returns
-------
"""
# Library.
from pyramid.arima import auto_arima
from pyramid.arima.arima import ARIMA
# Compute auto_arima.
results = auto_arima(**kwargs)
# Return a single PyramidWrapper object.
if isinstance(results, ARIMA):
return [self.from_instance(results)]
# Return an array of PyramidWrapper objects.
if isinstance(results, list):
return [PyramidWrapper().from_instance(a) for a in results]
def auto_fit_(self, init_p, init_q, max_p, max_q, m, init_P, d, D,
seasonal, stepwise):
"""
Tunning the parameters of ARIMA
"""
#Creating train and test sets
train_size = int(0.8 * len(self.timeseries))
train = self.timeseries[0:train_size]
stepwise_fit = auto_arima(
train,
start_p=init_p,
start_q=init_q,
max_p=max_p,
max_q=max_q,
m=m,
start_P=init_P,
seasonal=True,
d=d,
D=D,
trace=True,
error_action=
'ignore', # don't want to know if an order does not work
suppress_warnings=True, # don't want convergence warnings
stepwise=stepwise) # set to stepwise
return stepwise_fit.summary()
def auto_arimax(X, test_split=0.2):
# Do : data['Date'] = data['Date'].astype('datetime64[ns]')
# data.set_index("Date", inplace = True)
# Before sending data. The x axis labels wont get plotted if not done.
# plt.imsave() outside this function to save plot
test_samples = int(X.shape[0] * test_split)
train_data, test_data = X[:-test_samples], X[-test_samples:]
train_data.columns = ["Training Data"]
test_data.columns = ["Test Data"]
stepwise_model = auto_arima(X,
start_p=1,
start_q=1,
max_p=3,
max_q=3,
m=12,
start_P=0,
seasonal=True,
d=1,
D=1,
trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
stepwise_model.fit(train_data)
predictions = stepwise_model.predict(n_periods=len(test_data))
predictions = pd.DataFrame(predictions,
index=test_data.index,
columns=['Prediction'])
result = pd.concat([train_data, test_data, predictions], axis=1)
result.plot()
return rmse(np.array(test_data).flatten(), np.array(predictions).flatten())
def multi_process(i, df, date):
try:
df2 = df.loc[df['Stock Symbol'] == i].drop(columns='Stock Symbol')
df2.isnull().sum()
df2['Date'] = pd.to_datetime(df2['Date'],
format='%Y-%m-%d',
errors='coerce')
df2t = df2[df2['Date'] <= date].iloc[-713:, ]
df2v = df2[df2['Date'] > date].iloc[:1, ]
df2t = df2t.set_index('Date')
df2t_2 = df2t.diff().dropna()
df2v = df2v.set_index('Date')
model = auto_arima(df2t_2,
trace=False,
error_action='ignore',
suppress_warnings=True)
model.fit(df2t_2)
next_day_forecast = model.predict(
n_periods=1)[0] + df2t.iloc[-1:, ].values[0][0]
AIC = model.aic()
next_day_value = df2v.iloc[:1, ].values[0][0]
last_day_value = df2t.iloc[-1:, ].values[0][0]
l = [i, last_day_value, next_day_forecast, AIC, next_day_value]
return l
except:
print('Error with ', i)
def test_with_seasonality5():
# can we fit the same thing with an exogenous array of predictors?
# also make it stationary and make sure that works...
all_res = auto_arima(wineind,
start_p=1,
start_q=1,
max_p=2,
max_q=2,
m=12,
start_P=0,
seasonal=True,
n_jobs=1,
d=1,
D=None,
error_action='ignore',
suppress_warnings=True,
stationary=True,
random=True,
random_state=42,
return_valid_fits=True,
n_fits=5,
exogenous=rs.rand(wineind.shape[0], 4)) # only fit 2
# show it is a list
assert hasattr(all_res, '__iter__')
def test_the_r_src():
# this is the test the R code provides
fit = ARIMA(order=(2, 0, 1), trend='c', suppress_warnings=True).fit(abc)
# the R code's AIC = ~135
assert abs(135 - fit.aic()) < 1.0
# the R code's BIC = ~145
assert abs(145 - fit.bic()) < 1.0
# R's coefficients:
# ar1 ar2 ma1 mean
# -0.6515 -0.2449 0.8012 5.0370
# note that statsmodels' mean is on the front, not the end.
params = fit.params()
assert_almost_equal(params,
np.array([5.0370, -0.6515, -0.2449, 0.8012]),
decimal=2)
# > fit = forecast::auto.arima(abc, max.p=5, max.d=5, max.q=5, max.order=100, stepwise=F)
fit = auto_arima(abc,
max_p=5,
max_d=5,
max_q=5,
max_order=100,
seasonal=False,
trend='c',
suppress_warnings=True,
error_action='ignore')
# this differs from the R fit with a slightly higher AIC...
assert abs(137 - fit.aic()) < 1.0 # R's is 135.28
def train(self, base_dataset, start_date=None, end_date=None):
"""
Trains the model.
:param Dataset base_dataset: The dataset used to extract the training set
in accordance with the date range.
:param start_date: The minimum date for the records used in the training set.
:type start_date: datetime.datetime or None
:param end_date: The maximum date for the records used in the training set.
:type end_date: datetime.datetime or None
"""
if start_date is not None and end_date is not None and start_date > end_date:
raise ValueError('Invalid training date range')
training_set = base_dataset.getDataframe(ticker_symbol=self.ticker_symbol,
from_date=start_date, to_date=end_date)
self.training_start = min(training_set.date)
self.training_end = max(training_set.date)
# Pre-Processing
# ARIMA only receives as sequence of value in the training,
# then only the sequence of closing prices is needed
training_set = training_set.close
self.model = auto_arima(training_set,
start_p=1, start_q=1, max_p=3, max_q=3, m=12, start_P=0,
seasonal=True, d=1, D=1, trace=True,
error_action='ignore', suppress_warnings=True)
self.model.fit(training_set)
def calculate_prediction_data_arima(past_data, date_to):
result = dict()
for col in past_data.keys():
if len(list(past_data[col].keys())) > 0:
result[col] = dict()
factor = pd.DataFrame.from_dict(past_data[col], orient='index')
factor.index = pd.to_datetime(factor.index)
stepwise_model = auto_arima(factor,
start_p=1,
start_q=1,
max_p=3,
max_q=3,
m=12,
start_P=0,
seasonal=True,
d=1,
D=1,
trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
stepwise_model.fit(factor)
future = stepwise_model.predict(
n_periods=util.get_prediction_periods(date_to))
future_dt = util.get_prediction_datetimes_dt(
datetime.now(), date_to)
for i in range(len(future_dt)):
result[col][future_dt[i].strftime(
consts.DATE_FORMAT)] = future[i]
result = filter_unrealistic_values(result)
return result
def fit(self): #训练模型
# self.fittedModel = auto_arima(y=self.ts_train)
self.fittedModel = auto_arima(
y=self.ts_train,
start_p=self.start_p,
start_q=self.start_q,
max_p=self.max_p,
max_q=self.max_q,
max_d=self.max_d,
max_order=self.max_order,
seasonal=self.seasonal,
m=self.m,
# start_P=2,
# start_Q=0,
# D=1,
# max_Q=0,
maxiter=1000,
test='adf',
trace=False,
error_action=
'ignore', # don't want to know if an order does not work
suppress_warnings=True, # don't want convergence warnings
stepwise=self.stepwise,
information_criterion='bic',
njob=-1)
def ArimaxKLX(trainDF, predDF):
TrainExogenous = {'Month' : trainDF['MonthYear'].dt.month, 'Year' : trainDF['MonthYear'].dt.year}
TrainExogenousDF = pd.DataFrame(TrainExogenous)
TestExogenous = {'Month' : predDF['MonthYear'].dt.month, 'Year' : predDF['MonthYear'].dt.year}
TestExogenousDF = pd.DataFrame(TestExogenous)
arimax_fit = auto_arima(trainDF['OrderQuantity'],
exogenous=TrainExogenousDF,
start_p=0,
start_q=0,
max_p=6,
max_d=2,
max_q=6,
start_P=0,
start_Q=0,
max_P=6,
max_D=2,
max_Q=6,
max_order=6,
seasonal=True,
stationary=False,
information_criterion='aic',
stepwise=False,
trace=False,
test='adf',
seasonal_test='ocsb',
error_action='ignore',
suppress_warnings=True,
enforce_stationarity=False)
ArimaxForecast = arimax_fit.predict(len(predDF), exogenous=TestExogenousDF)
ArimaxForecast = np.round(ArimaxForecast, 0)
ArimaxForecast = np.clip(ArimaxForecast, 0, np.max(ArimaxForecast))
return ArimaxForecast
def predicting(data, pltname):
data = data[['_time', 'cycles']]
original = data['cycles']
n = len(data)
#print("length of datta",n)
forecast_out = int(math.ceil(0.2 * (n)))
#print("forecat_out",forecast_out)
data['label'] = data['cycles'].shift(-forecast_out)
data.dropna(inplace=True)
original = data['cycles']
original = original.to_frame(name='cycles')
d1 = data[
'label'] #this step changesa dataframe object into that of a series.series object
d1 = d1.to_frame(
name='label') #thus need to convert it back into a dataframe object
data = data['label']
data = data.to_frame(name='label')
#print("last value maam",data.iat[len(data)-1,0])
#divide into train and validation set
#train = data[:int(0.8*(len(data)))]
#test = data[int(0.8*(len(data))):]
model = auto_arima(data,
trace=True,
start_p=0,
start_q=0,
start_P=0,
start_Q=0,
max_p=3,
max_q=3,
max_P=3,
max_Q=3,
seasonal=True,
stepwise=False,
suppress_warnings=True,
D=1,
max_D=10,
error_action='ignore',
approximation=False)
#change 3 to 10
#fitting model
model.fit(data)
#print(model.summary())
y_pred = model.predict(n_periods=forecast_out)
y_pred = pd.DataFrame(y_pred, columns=['label'])
#print("first element",y_pred.iat[0,0])
conn = pd.concat([d1, y_pred], axis=0)
n = conn.size - forecast_out
plt.figure(0)
plt.plot(original[:n], conn[:n], 'y')
diff = original.iat[len(original) - 1,
0] - original.iat[len(original) - forecast_out, 0]
plt.plot(original[-forecast_out:] + diff, conn[-forecast_out:], 'r')
plt.savefig("/Users/Arunima_Sharma/Desktop/py/flask/static/" + pltname)
plt.show()
return y_pred
def test_with_seasonality3():
# show we can estimate D even when it's not there...
auto_arima(
wineind,
start_p=1,
start_q=1,
max_p=2,
max_q=2,
m=12,
start_P=0,
seasonal=True,
d=1,
D=None,
error_action='ignore',
suppress_warnings=True,
trace=True, # get the coverage on trace
random_state=42,
stepwise=True)
def forecast_traces(balances, weeks=52) -> List[Scatter]:
"""
Forecast next weeks based on balance history with ARIMA
"""
last_balance_date = balances.index.max()
last_balance = balances.loc[last_balance_date]
X_forecast = list(pd.date_range(last_balance_date, periods=weeks,
freq='W'))
model = auto_arima(balances,
trend=[1, 1],
error_action='ignore',
suppress_warnings=True)
forecast = model.predict(n_periods=weeks)
# error estimation
model_error = np.std(model.resid())
sampling_error = np.sqrt(balances.var() / len(balances))
forecast_error = 2 * np.sqrt(model_error**2 + sampling_error**2)
forecast_upper = forecast + forecast_error
forecast_lower = forecast - forecast_error
bad_forecast = np.min(forecast) < balances.loc[last_balance_date]
balances_forecast = Scatter(
name='forecast',
x=X_forecast,
y=[last_balance] + list(forecast),
mode='lines',
line={
'dash': 'dash',
'color': color('EXP') if bad_forecast else color('INC')
},
opacity=0.8)
balances_forecast_upper = Scatter(name='',
x=X_forecast,
y=[last_balance] + list(forecast_upper),
fill='tonexty',
fillcolor=(color(
'EXP' if bad_forecast else 'INC',
alpha=0.2)),
line={'color': 'transparent'},
showlegend=False)
balances_forecast_lower = Scatter(name='',
x=X_forecast,
y=[last_balance] + list(forecast_lower),
fill='tozeroy',
fillcolor='#8881',
line={'color': 'transparent'},
showlegend=False)
return [
balances_forecast_lower,
balances_forecast_upper,
balances_forecast,
]
def test_force_polynomial_error():
x = np.array([1, 2, 3, 4, 5, 6])
d = 2
xreg = None
with pytest.raises(ValueError) as ve:
auto_arima(x, d=d, D=0, seasonal=False, exogenous=xreg)
assert 'simple polynomial' in str(ve), str(ve)
# but it should pass when xreg is not none
xreg = rs.rand(x.shape[0], 2)
_ = auto_arima(x,
d=d,
D=0,
seasonal=False,
exogenous=xreg,
error_action='ignore',
suppress_warnings=True)
def fitFunc(yvals, ndays):
model = auto_arima(yvals, trace=True, error_action='ignore', suppress_warnings=True)
model.fit(yvals)
n_periods = ndays
forecast = model.predict(n_periods)
oned = range(len(yvals), len(yvals) + len(forecast))
oned2 = range(0, len(yvals))
return [forecast[-1], oned2, yvals, oned, forecast]
def build_arima_model(data, p=0, d=0, q=0, predict=False, show_qqplot=False):
forest_date = [
x.strftime('%Y-%m-%d')
for x in list(pd.date_range(start='2014-09-01', end='2014-09-30'))
]
if p == 0 and d == 0 and q == 0:
model = auto_arima(data)
model.fit(data)
else:
model = sm.tsa.ARIMA(data, order=(p, d, q))
results = model.fit()
# 滚动预测
# forest = []
# for i in range(0, len(forest_date)-7):
# if i % 7 == 0:
# temp = []
# temp += results.forecast(steps=7)[0].tolist()
# temp_pd = pd.Series(temp, index=pd.to_datetime(forest_date[i:i + 7]))
# img = img.append(temp_pd)
# forest += temp
# model = None
# model = sm.tsa.ARIMA(img, order=(p, d, q))
# results = model.fit()
# forest += results.forecast(steps=2)[0].tolist()
if show_qqplot:
resid = results.resid.values
for i in range(0, len(resid)):
resid[i] = round(resid[i], 8)
qqplot(resid)
# 模型历史数据预测
if predict:
if p == 0 and d == 0 and q == 0:
predict = model.predict_in_sample(start=0, end=151, dynamic=False)
else:
predict = results.predict(start=str('2014-04-03'),
end=str('2014-08-30'),
dynamic=False)
predict_date = [
x.strftime('%Y-%m-%d')
for x in list(pd.date_range(start='2014-04-01', end='2014-08-30'))
]
predict_df = pd.DataFrame({'forest': predict},
index=pd.to_datetime(predict_date))
fig, ax = plt.subplots(figsize=(12, 8))
ax = data.plot(ax=ax)
ax = predict_df.plot(ax=ax)
if p == 0 and d == 0 and q == 0:
forest = model.predict(n_periods=30)
else:
forest = results.forecast(steps=30)[0]
forest_df = pd.DataFrame({'forest': forest},
index=pd.to_datetime(forest_date))
fig, ax = plt.subplots(figsize=(12, 8))
ax = data.plot(ax=ax)
ax = forest_df.plot(ax=ax)
plt.show()
return forest_df
def test_many_orders():
# show that auto-arima can't fit this data for some reason...
lam = 0.5
lynx_bc = ((lynx**lam) - 1) / lam
failed = False
try:
auto_arima(lynx_bc,
start_p=1,
start_q=1,
d=0,
max_p=5,
max_q=5,
n_jobs=-1,
suppress_warnings=True,
maxiter=10) # shorter max iter
except ValueError:
failed = True
assert failed
def arimamodelling(timeseries):
automodel = auto_arima(timeseries,
start_p=1,
start_q=1,
max_p=10,
max_q=10,
trace=True,
error_action='ignore',
suppress_warnings=True)
return automodel
data1 = data1.iloc[:, 1:37]
no_of_column=len(data1.columns)
forecasted=pd.DataFrame()
for row in data1.iloc[0:20,:].iterrows():
index, data = row
if(np.sum(data[no_of_column-3:no_of_column])==0):
print("Next")
continue
data=data.replace(0,np.median(data))
stepwise_model = auto_arima(data, start_p=1, start_q=1,
max_p=3, max_q=3, m=12,
start_P=0, seasonal=True,
d=1, D=1, trace=True,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
print(index)
print(row)
stepwise_model.aic()
future_forecast = stepwise_model.predict(n_periods=12)
forecasted=forecasted.append(pd.Series(future_forecast),ignore_index=True)
# Example
plt.plot(data)
plt.plot(future_forecast)
print(future_forecast)