def arima_predict(train_dat, n_predictions, p=2, d=0, q=0):
arima = ARIMA(np.array(train_dat).astype(np.float), [p, d, q])
diffed_logged_results = arima.fit(trend='c', disp=False)
preds = diffed_logged_results.predict(len(train_dat),
len(train_dat) + n_predictions - 1,
exog=None, dynamic=False)
return preds
def ARIMA_forcast2(self):
# this approach forecast 1 data pt at a time, then add the new forecast datapoint to the training data
# then repeat
import warnings
warnings.filterwarnings('ignore')
# test without taking log of data
# using rolling avg
y = vr_df2_ts.values
train = vr_df2_ts.values[286:574]
prediction = list()
for t in range(288):
modelY = ARIMA(y, order=(1,1,1))
results = modelY.fit(disp=-1)
out = results.forecast()
yhat = out[0]
prediction.append(yhat)
y = np.append(y,train[t])
forecast = pd.Series(prediction,index=pd.date_range(start='2017-02-09 00:00:00', periods=288,freq='5min'))
exog = vr_df2_ts.iloc[286:574]
exog.set_index(pd.date_range(start='2017-02-09 00:00:00', periods=288,freq='5min'),inplace=True)
plt.plot(vr_df2_ts)
plt.plot(exog,'g')
plt.plot(forecast,'r')
def forecast_by_cluster(self, hold_out_n, n_ahead, order, exog):
dfit = self.ds_agg_by_c
efit = efor = None
if hold_out_n > 0:
# hold out validation required
dfit = dfit[:-hold_out_n]
if (exog is not None):
efit = exog[:-hold_out_n]
efor = exog[-hold_out_n:]
else:
if (exog is not None):
efit = exog[:-n_ahead]
efor = exog[-n_ahead:]
ds_c_for = np.zeros((n_ahead, self.n_clusters))
for c in tqdm(range(self.n_clusters)):
cdfit = dfit[:,c]
if sum(cdfit) == 0:
ds_c_for[:,c] = 0
continue
m = ARIMA(cdfit, exog = efit, order = order)
mf = m.fit()
f = mf.forecast(n_ahead, exog = efor, alpha = .95)[0]
ds_c_for[:,c] = f
self.ds_c_for = ds_c_for
def arimamodel(ts):
ts_log, ts_log_diff = trend(ts)
model = ARIMA(ts_log, order = (2,1,2))
result_ARIMA = model.fit(disp = -1)
m = ARIMA(ts, order = (2,1,2)).fit()
arimares = ARMAResults(m, params = '')
pre = arimares.forcast(steps = 60)
# pre = m.predict('20150901', '20151230', dynamic = True)
print pre
# prediction back to the original scale
predictions_ARIMA = backorg(result_ARIMA, ts_log)
plt.plot(predictions_ARIMA)
# print (predictions_ARIMA - ts)[40:80]
plt.plot(ts, color = 'red')
# plt.plot(ts_log_diff)
# plt.plot(result_ARIMA.fittedvalues, color = 'red')
plt.title('RSS: %.4F' % np.sum((result_ARIMA.fittedvalues - ts_log_diff)**2))
plt.show()
def get_grouped_data(self, forecast=False):
cdf = self.cumulative_sum()
gdf = self.group_by('M')
if cdf.shape[0] > gdf.shape[0]:
df = cdf.to_frame()
df.columns = ['cumulative sum']
df['total added'] = gdf.to_frame()['event']
else:
df = gdf.to_frame()
df.columns = ['total added']
df['cumulative sum'] = cdf.to_frame()['event']
if forecast:
mtotals = pd.to_numeric(df['cumulative sum'], downcast='float')
model = ARIMA(mtotals, order=(10,1,0))
model_fit = model.fit(disp=0)
forecast = model_fit.forecast(steps=12)
dates = pd.date_range('2017-04-30', '2018-06-01', freq='M')
records = zip([x.to_datetime() for x in dates], forecast[0])
ndf = pd.DataFrame.from_records(records)
ndf.columns = ['date', 'forecast']
ndf.set_index(['date'], inplace=True)
df = pd.concat([df, ndf], axis=1)
return df
def mamodel(ts):
ts_log, ts_log_diff = trend(ts)
model = ARIMA(ts_log, order = (0,1,1))
result_MA = model.fit(disp = -1)
plt.plot(ts_log_diff)
plt.plot(result_MA.fittedvalues, color = 'red')
plt.title('RSS: %.4F' % np.sum((result_MA.fittedvalues - ts_log_diff)**2))
plt.show(block = False)
def armodel(ts):
ts_log, ts_log_diff = trend(ts)
model = ARIMA(ts_log, order = (1,1,0))
result_AR = model.fit(disp = -1)
plt.plot(ts_log_diff)
plt.plot(result_AR.fittedvalues, color = 'red')
# pdb.set_trace()
plt.title('RSS: %.4F' % np.sum((result_AR.fittedvalues - ts_log_diff)**2))
plt.show(block = False)
def ARIMA_fit(self):
# order=(p,d,q) AR and MA can also be modeled separately by enter 0 for either p or q
model = ARIMA(ts_log, order=(5,1,5))
self.results_ARIMA = model.fit(disp=-1)
print(results_ARIMA.summary())
plt.plot(ts_log_diff)
plt.plot(results_ARIMA.fittedvalues, color='r')
plt.title('RSS: %.4f'% sum((results_ARIMA.fittedvalues-ts_log_diff['in_tpkts'])**2))
def fit(self): if len(self.df) < self.t_window: return None model = ARIMA(self.df, order=(2, 1, 1)) results_ARIMA = model.fit(disp=-1) forecast = results_ARIMA.predict(start = self.t_window, end= self.t_window+2, dynamic= True) forecast = forecast.cumsum() predictions_ARIMA_log = pd.Series(self.df.ix[self.t_window-1], index=forecast.index) predictions_ARIMA_log = predictions_ARIMA_log.add(forecast,fill_value=0) predictions_ARIMA = np.exp(predictions_ARIMA_log) #print self.df return predictions_ARIMA
def ARIMA_fun( data ):
lag_pacf = pacf( data, nlags=20, method='ols' )
lag_acf, ci2, Q = acf( data, nlags=20 , qstat=True, unbiased=True)
model = ARIMA(orig_data, order=(1, 1, int(ci2[0]) ) )
results_ARIMA = model.fit(disp=-1)
plt.subplot(121)
plt.plot( data )
plt.plot(results_ARIMA.fittedvalues)
#plt.show()
return results_ARIMA.fittedvalues
def objfunc(order, *params):
series = params
try:
mod = ARIMA(series, order, exog=None)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
res = mod.fit(disp=0, solver='bfgs', maxiter=5000)
except:
return float('inf')
if math.isnan(res.aic):
return float('inf')
return res.aic
def pridictNextNdays(self,train):
timeSerize = train[self.selected]
timeSerize = timeSerize[self.start_train:self.end_train]
model = ARIMA(timeSerize, order=(self.p,self.d,self.q), freq='D') # build a model
fitting = model.fit(disp=False)
forecast, fcasterr, conf_int = fitting.forecast(steps=self.next_ndays, alpha=.05)
# params = fitting.params
# residuals = fitting.resid
# p = fitting.k_ar
# q = fitting.k_ma
# k_exog = fitting.k_exog
# k_trend = fitting.k_trend
# forecast = _arma_predict_out_of_sample(params,self.next_ndays,residuals, p, q, k_trend, k_exog, endog=timeSerize, exog=None, start=len(timeSerize))
return forecast
def testArima(self,train):
realSerize = train[self.selected]
timeSerize = realSerize[self.start_train:self.end_train]
realData = train[self.selected][self.end_train:self.next_ndays]
model = ARIMA(timeSerize, order=(self.p,self.d, self.q)) # build a model
fitting = model.fit(disp=False)
forecast, fcasterr, conf_int = fitting.forecast(steps=self.next_ndays, alpha=.05)
# params = fitting.params
# residuals = fitting.resid
# p = fitting.k_ar
# q = fitting.k_ma
# k_exog = fitting.k_exog
# k_trend = fitting.k_trend
# forecast = _arma_predict_out_of_sample(params,self.next_ndays,residuals, p, q, k_trend, k_exog, endog=timeSerize, exog=None, start=len(timeSerize))
return {'real':list(realSerize)[self.end_train:self.end_train+self.next_ndays],'pridiction':forecast}
def predict_arima_next_days(self, item):
ts = df_train[item]
ts = ts.sort_index() # sorting index Date
ts_last_day = ts[self.fc] # real last data
ts = ts[0:self.fc] # index 0 until last data - 1
model = ARIMA(ts, order=(self.p, self.d, self.q)) # build a model
fitting = model.fit(disp=False)
# n_days forecasting
forecast, fcasterr, conf_int = fitting.forecast(steps=self.n_days, alpha=.05)
# ts: history until 1 day before self.fc
# ts[self.fc]: last day
# forecast: 1 day forecast (time equalto ts[self.fc])
return ts, ts_last_day, forecast
def arima(ts, forecast_window):
logger.info(ts)
start = int(ts.count() - 1)
end = int(start + forecast_window)
ts_log = np.log(ts)
model = ARIMA(ts_log, order=(0, 1, 2))
results = model.fit(disp=-1)
prediction = results.predict(start=start, end=end, dynamic=True)
future = pd.Series(prediction, copy=True)
cumsum = future.cumsum()
prediction_future = future.add(ts_log.ix[-1])
prediction_future = prediction_future.add(cumsum)
ts_future = np.exp(prediction_future)
return ts_future
class ARIMAModelResult:
def __init__(self, autoregressive_periods, integrated_order, moving_average_model_periods, training_data, test):
self.autoregressive_periods = autoregressive_periods
self.integrated_order = integrated_order
self.moving_average_model_periods = moving_average_model_periods
self.model = ARIMA(training_data, order=(
self.autoregressive_periods,
self.integrated_order,
self.moving_average_model_periods
)
)
self.fit = self.model.fit()
self.aic = self.fit.aic
self.predictions = self.fit.forecast(steps=len(test))[0]
self.model_fitness = mean_squared_error(test, self.predictions)
def __eq__(self, other):
return self.model_fitness == other.model_fitness
def __lt__(self, other):
return self.model_fitness < other.model_fitness
def __gt__(self, other):
return self.model_fitness > other.model_fitness
def __str__(self):
return "Autoregressive periods: {}\nIntegraded Order: {}\nMoving Average Model Periods: {}\n Predictions: {}\nMSE: {}".format(
self.autoregressive_periods,
self.integrated_order,
self.moving_average_model_periods,
self.predictions,
self.model_fitness
)
def predictFutureProfit(df, forward):
results = {}
for asset in get_assets(df):
ts = df[asset]
ts_log = np.log(ts)
model = ARIMA(ts_log, order=(1, 1, 0))
results_ARIMA = model.fit(disp=-1)
predictions_diff = results_ARIMA.predict(2, len(ts.index)-1, dynamic=True)
predictions_diff_cumsum = predictions_diff.cumsum()
predictions_log = pd.Series(ts_log.ix[0], index=ts_log.index)
predictions_log = predictions_log.add(predictions_diff_cumsum,fill_value=0)
predictions = np.exp(predictions_log)
results[asset] = predictions[-1]
return results
def arima(self):
kl = self.get_kline()
cp = self.get_close_price(kl)
date = self.get_date(kl)
#t = datetime.fromtimestamp(date[-1].timestamp()+24*60*60)
t = date[-1] + timedelta(days=int(self.day_history/5)) #days seconds ...
print("predict date:", date[-1],"--->", t)
dta = pd.Series(cp, index=date)
print(dta)
model=ARIMA(dta,order=(4,1,3)) #P D Q
result=model.fit()
pred=result.predict( date[-10], t,dynamic=True,typ='levels')
plt.figure(figsize=(12,8))
plt.plot(dta, 'ro-')
plt.xticks(rotation=45)
plt.plot(pred, 'go-')
plt.show()
def fitArima(ts):
import statsmodels.api as sm
logged_ts = np.log(ts)
diffed_logged_ts = (logged_ts - logged_ts.shift(7))[7:]
p = 0
d = 1
q = 1
arima = ARIMA(diffed_logged_ts, [p, d, q], exog=None, freq='D', missing='none')
diffed_logged_results = arima.fit(trend='c', disp=False)
predicted_diffed_logged = diffed_logged_results.predict(exog=None, dynamic=False)
#a=pd.date_range(diffed_logged_ts.index[1], periods=90, freq='D')
predicted_diffed_logged_ts = pd.Series(predicted_diffed_logged, index=diffed_logged_ts.index[d:])
predicted_diffed_logged_ts = np.exp(logged_ts.shift(7) + diffed_logged_ts.shift(d) + predicted_diffed_logged_ts)
concatenated = pd.concat([ts, predicted_diffed_logged_ts], axis=1, keys=['original', 'predicted'])
#a= concatenated
#a.plot()
#plt.show()
return concatenated
def programmer_5():
discfile = "data/discdata_processed.xls"
# 残差延迟个数
lagnum = 12
data = pd.read_excel(discfile, index_col="COLLECTTIME")
data = data.iloc[:len(data) - 5]
xdata = data["CWXT_DB:184:D:\\"]
# 训练模型并预测，计算残差
arima = ARIMA(xdata, (0, 1, 1)).fit()
xdata_pred = arima.predict(typ="levels")
pred_error = (xdata_pred - xdata).dropna()
lb, p = acorr_ljungbox(pred_error, lags=lagnum)
h = (p < 0.05).sum()
if h > 0:
print(u"模型ARIMA（0,1,1)不符合白噪声检验")
else:
print(u"模型ARIMA（0,1,1)符合白噪声检验")
print(lb)
def run_arima(self):#use current build
'''
DEPRECATED:
Primarily used for testing/debugging.
Runs statsmodels ARIMA.
'''
self.xts = self.X_train.set_index('date')
self.yts = self.y_train.set_index('date')
self.yts.astype('float', inplace=True)
self.arimod = ARIMA(endog = self.yts, order = (2,1,2))#, exog=self.xts)
self.aresults = self.arimod.fit()
def arima_model(accounts):
"""Fit ARIMA models for each account"""
# Model each account
account_models = {}
for account_type, account in accounts:
account_data = accounts[(account_type, account)]
account_data.name = account
# ARIMA model order is unknown, so find the highest order that can be fit
order = 0
modeled = False
while not modeled and order < len(ARIMA_ORDERS):
try:
model = ARIMA(account_data, order=ARIMA_ORDERS[order])
results = model.fit()
modeled = True
account_models[(account_type, account)] = results
except (ValueError, np.linalg.LinAlgError):
order += 1
return account_models
def ARIMA_forcast3(self):
# load dataset
series = pd.Series(vr_df['ACTIVE_FLOWS'][0:7000])
# seasonal difference
X = series.values
cycle = 288 #2016
differenced = difference(X, cycle)
# fit model
model = ARIMA(differenced, order=(1,1,1))
model_fit = model.fit(disp=0)
# multi-step out-of-sample forecast
forecast = model_fit.forecast(steps=2016)[0]
# invert the differenced forecast to something usable
history = [x for x in X]
step = 1
forecast_values = []
for yhat in forecast:
inverted = inverse_difference(history, yhat, cycle)
#print('Day %d: %f' % (day, inverted))
forecast_values.append(inverted)
history.append(vr_df['ACTIVE_FLOWS'][7000+step-1])
step += 1
def ARIMA_forecast4(self):
# parameters
num_train_init = 7318
num_forecast = 12 #one day = 288 data points
cycle = 288 #for a total 288 samples per day
startdate = vr_df.index[num_train]
field = 'DELETED_FLOWS'
# array of predicted values
forecast_values = []
for i in range(0,int(len(vr_df)/num_forecast)):
# check array for out of bound
num_train_current = i*num_forecast+num_train_init
if ((num_train_current) > len(vr_df)):
break
# load dataset
series = pd.Series(vr_df[field][0:num_train_current])
# Make data stationary: seasonal difference
X = series.values
differenced = difference(X, cycle)
# fit model
model = ARIMA(differenced, order=(1,1,1))
model_fit = model.fit(disp=0)
# multi-step out-of-sample forecast
forecast = model_fit.forecast(steps=num_forecast)[0]
# invert the differenced forecast to something usable
history = [x for x in X]
step = 1
for yhat in forecast:
inverted = inverse_difference(history, yhat, cycle)
forecast_values.append(inverted)
#append actual data
try:
history.append(vr_df[field][num_train_current+step-1])
except:
# reached the end of actual data array, use forecasted values to estimate
history.append(inverted)
step += 1
def __init__(self, autoregressive_periods, integrated_order, moving_average_model_periods, training_data, test):
self.autoregressive_periods = autoregressive_periods
self.integrated_order = integrated_order
self.moving_average_model_periods = moving_average_model_periods
self.model = ARIMA(training_data, order=(
self.autoregressive_periods,
self.integrated_order,
self.moving_average_model_periods
)
)
self.fit = self.model.fit()
self.aic = self.fit.aic
self.predictions = self.fit.forecast(steps=len(test))[0]
self.model_fitness = mean_squared_error(test, self.predictions)
def getLikelihood(endog,exog, order = None,n_forecasted_data=1):
'''
train_en = endog[:predict_start-1]
test_en = endog[predict_start:]
print train_en
print test_en
train_ex = exog[:predict_start-1]
test_ex = exog[predict_start:]
'''
# Automatically determine values of orders
if order is None:
from scipy.optimize import brute
grid = (slice(1, 3, 1), slice(1, 3, 1),slice(0, 3, 1))
print "############################################"
print endog
print "############################################"
try:
order = brute(objfunc, grid, args=(exog, endog), finish=None)
order = order.astype(int)
except :
order = [1,1,3]
# Model fits given data (endog) with optimized order
print "*********************************************"
print "Choose order of ",
print order
print "*********************************************"
model = ARIMA(endog,order).fit(full_output=False,disp=False)
# 1st element of array x is the forecasted data.
x = model.forecast(n_forecasted_data)
return x[0]
def previsao_matematica(reservatId, data):
seriesArray = Series.from_array(predict_info.getSeries(reservatId, data))
seriesValues = seriesArray.values
mathDict = {'calculado': False, 'volumes': [], 'dias': 0}
#if isNonStationary(seriesValues) == True:
days_in_year = 1
differenced = predict_info.difference(seriesValues, days_in_year)
# fit model
model = ARIMA(differenced, order=(1,0,1))
model_fit = model.fit(disp = -1)
# multi-step out-of-sample forecast
forecast = model_fit.forecast(steps=180)[0]
# invert the differenced forecast to something usable
mathDict['calculado'] = True
history = [x for x in seriesValues]
for yhat in forecast:
inverted = predict_info.inverse_difference(history, yhat, days_in_year)
history.append(inverted)
if inverted >= 0.0:
mathDict['volumes'].append("%.4f" % round((inverted), 4))
mathDict['dias'] = mathDict['dias'] + 1
return mathDict
def get_arima_predictions(y, train_subset, order = [1,0,0], X = None):
if X == None:
arima = ARIMA(y[train_subset], order = order).fit()
predictions = arima.predict()
else:
arima = ARIMA(y[train_subset], order = order,
exog = X[train_subset,:]).fit()
predictions = arima.predict(exog = X[train_subset,:])
for i in range(max(train_subset)+1,len(y)):
if X == None:
arima = ARIMA(y[0:i], order = order).fit()
predictions = np.append(predictions,
arima.predict(0, len(y) + i)[-1])
else:
arima = ARIMA(y[0:i], order = order, exog = X[0:i,:]).fit()
predictions = np.append(predictions,
arima.predict(0, len(y) + i,
exog = X[0:i+1,:])[-1])
return predictions
# Initialize local variable for time series
trailer_series = subtype_result['Trailer']
# trailer_series = subtype_result_day['Trailer']
# trailer_series = subtype_result['Trailer'].resample('MS').sum()
X = trailer_series.values
train, test = X[0:-52], X[-52:]
history = [x for x in train]
# print(history)
predictions = list()
for t in range(len(test)):
# fit model
model = ARIMA(history, order=(4, 1, 0))
model_fit = model.fit(disp=False, trend='c')
# single step forecast
yhat = model_fit.forecast()[0]
predictions.append(yhat)
history.append(test[t])
# evaluate forecasts
rmse = sqrt(mean_squared_error(test, predictions))
print('Test RMSE: %.3f' % rmse)
print(model_fit.summary())
# model_fit.plot_diagnostics(figsize=(16, 8))
# Plot The Forecast
plt.plot(test, color='#ff6832')
def main():
# parse arguments
args = add_args()
# set the level of logger
logger.setLevel(logging.DEBUG)
if not args.verbose:
logger.setLevel(logging.INFO)
logger.debug("--------DEBUG enviroment start---------")
# show the hyperparameters
logger.info("---------hyperparameter setting---------")
logger.info(args)
# set the random seed
np.random.seed(args.seed)
# data fetching
logger.info("-------------Data fetching-------------")
tickers = \
[
("TSLA", "yahoo"), # 0, TESLA Stock
]
# check if data range is legal.
if args.month <= 0 or args.month > 24:
logger.warning("The data range is illegal. Turn to use default 3")
args.month = 3
tsla_df = data_loader(
tickers, args.month)[0] # get dataframes from "yahoo" finance.
tsla_close = tsla_df["Close"].resample(
'D').ffill() # fullfill the time series.
# data cleaning
logger.info("-------------Data cleaning-------------")
if np.sum(tsla_close.isnull()) > 0:
logger.debug(
"The time series contain missing values & we use interpolation to resolve this issue"
)
tsla_close = tsla_close.interpolate(method='polynomial',
order=2,
limit_direction='forward',
axis=0)
# Then, if there is still some missing values, we simply drop this value.abs
tsla_close = tsla_close.dropna()
logger.debug(tsla_close)
# plot the graph describe tsla close
if args.plot:
fig = plt.gcf()
fig.set_size_inches(18.5, 10.5)
plt.plot(tsla_close, label="Series")
plt.plot(tsla_close.rolling(int(.05 * len(tsla_close))).mean(),
'--',
label="Rolling mean")
plt.plot(tsla_close.rolling(int(.05 * len(tsla_close))).std(),
":",
label="Rolling Std")
plt.legend(loc="best")
plt.savefig("tesla_description.png")
# if log transformation
if args.log:
tsla_close = tsla_close.apply(np.log) # log transformation
# estimate the forecastability of a time series:
# Approximate entropy is a technique used to quantify the amount of regularity and the unpredictability of fluctuations over time-series data.
# Smaller values indicates that the data is more regular and predictable.
logger.info("The approximate entropy: " + str(
app_entropy(U=np.array(tsla_close),
r=0.2 * np.std(np.array(tsla_close)))))
# data splitting
logger.info("-------------Data splitting------------")
# check if split_ratio legal.
if args.split_ratio > 1 or round(len(tsla_close) * args.split_ratio) <= 0:
logger.warning("Splitting ratio is illegal. Turn to use default 0.7")
args.split_ratio = 0.7
train = tsla_close[0:round(len(tsla_close) * args.split_ratio)]
test = tsla_close[round(len(tsla_close) * args.split_ratio):]
# time serise decomposition
logger.info("-------------decomposition-------------")
# check if period is legal.
if args.period < 2:
logger.warning("Seasonal period is illegal. Turn to use default 7.")
args.period = 7
trend, seasonal, residual = decompose(train, args.period, args.plot)
# difference
logger.debug("-----------------Diff-----------------")
trend_diff, trend_diff_counts = diff(trend, args.plot, "trend", args.diff)
logger.debug("trend diff counts: " + str(trend_diff_counts))
residual_diff, residual_diff_counts = diff(residual, args.plot, "residual",
args.diff)
logger.debug("residual diff counts: " + str(residual_diff_counts))
# ARIMA model
logger.info("-----------ARIMA construction----------")
trend_model_fit, trend_model_order = ARIMA_model(trend_diff, args,
"trend_diff")
logger.info("Trend model parameters: " + str(
tuple([trend_model_order[0], trend_diff_counts, trend_model_order[1]]))
)
residual_model_fit, residual_model_order = ARIMA_model(
residual_diff, args, "residual_diff")
logger.info("Residual model parameters: " + str(
tuple([
residual_model_order[0], residual_diff_counts,
residual_model_order[1]
])))
# model summary
try:
logger.debug("---------trend model summary----------")
logger.debug(trend_model_fit.summary())
except:
logger.warning("Error occurs in summary, simply skip")
pass
try:
logger.debug("---------resid model summary----------")
logger.debug(residual_model_fit.summary())
except:
logger.warning("Error occurs in summary, simply skip")
pass
if args.plot:
# residual plots of trend model
trend_model_fit.resid.plot()
plt.savefig("resid_plt_trend.png")
plt.close()
trend_model_fit.resid.plot(kind='kde')
plt.savefig("kde_resid_plt_trend.png")
plt.close()
# residual plots of residual model
residual_model_fit.resid.plot()
plt.savefig("resid_plt_residual.png")
plt.close()
residual_model_fit.resid.plot(kind='kde')
plt.savefig("kde_resid_plt_residual.png")
plt.close()
logger.debug("-----trend model residual describe----")
logger.debug(trend_model_fit.resid.describe()) # describe the dataframe
logger.debug("-----resid model residual describe----")
logger.debug(residual_model_fit.resid.describe()) # describe the dataframe
# loss calculation
logger.info("-----------Loss calculation------------")
fit_seq = model_predict(trend_model_fit, residual_model_fit, trend,
residual, seasonal, trend_diff_counts,
residual_diff_counts, False, "", "", args.period)
if args.log:
fit_seq = np.exp(fit_seq)
train = train.apply(np.exp)
logger.debug(fit_seq)
# calculate training loss
training_loss = loss(fit_seq, np.array(train), args.loss)
logger.info("Training loss: " + str(training_loss))
# plot train and fitted values in one graph.
if args.plot:
plt.figure()
plt.plot(fit_seq, color='red', label='fit')
plt.plot(np.array(train), color='blue', label='train')
plt.legend(loc='best')
plt.savefig('fit_vs_train.png')
plt.close()
if list(test):
pred_seq = model_predict(trend_model_fit, residual_model_fit, trend,
residual, seasonal, trend_diff_counts,
residual_diff_counts, True,
str(test.index.tolist()[0]),
str(test.index.tolist()[-1]), args.period)
if args.log:
pred_seq = np.exp(pred_seq)
test = test.apply(np.exp)
logger.debug(pred_seq)
# calculate testing loss
testing_loss = loss(pred_seq, np.array(test), args.loss)
logger.info("Testing loss: " + str(testing_loss))
# plot test and predicted value in one graph.
if args.plot:
plt.figure()
plt.plot(pred_seq, color="red", label="pred")
plt.plot(np.array(test), color="blue", label="test")
plt.legend(loc="best")
plt.savefig("pred_vs_test.png")
plt.close()
# plot several models performance comparison on train set.
if args.plot:
logger.info("-----------Model Comparison------------")
plt.figure()
# actual value
plt.plot(np.array(train), color='blue', label="actual")
# auto-ARIMA with seasonal decompostion
plt.plot(fit_seq[1:],
color='green',
label='ARIMA with seasonal decomposition')
# simple auto-ARIMA
auto_arima_model_fit, _ = ARIMA_model(train, args, "auto_arima")
plt.plot(np.array(auto_arima_model_fit.fittedvalues),
color='yellow',
label='Auto ARIMA')
# auto-ARIMA with log transfromation.
auto_log_arima_fit, _ = ARIMA_model(train.apply(np.log), args,
"auto_arima")
plt.plot(np.array(auto_log_arima_fit.fittedvalues.apply(np.exp)),
color='brown',
label='Auto ARIMA with log')
# rolling mean
plt.plot(np.array(train.rolling(int(.05 * len(train))).mean()),
'--',
label="Rolling mean")
# ordinary arima
plt.plot(np.array(ARIMA(train, (1, 0, 1)).fit(disp=0).fittedvalues),
color="coral",
label="Ordinary ARIMA")
plt.legend(loc="best")
plt.xlabel("days from " +
str(train.index.tolist()[0]).replace(" 00:00:00", ""))
plt.ylabel("stock prices")
plt.title("Actual Stock Price Compared with Forecasted Stock Price")
plt.grid(True)
plt.tight_layout()
plt.savefig("model_comparison.png")
plt.close()
if list(test):
# calculate testing loss
loss_dict = dict()
# auto-ARIMA with seasonal decompostion
loss_dict["auto sarima"] = testing_loss
# simple auto-ARIMA
loss_dict["auto arima"] = loss(
np.array(
auto_arima_model_fit.predict(
start=str(test.index.tolist()[0]),
end=str(test.index.tolist()[-1]),
dynamic=True)), np.array(test), args.loss)
# auto-ARIMA with log transfromation.
loss_dict["auto arima log"] = loss(
np.array(
auto_log_arima_fit.predict(
start=str(test.index.tolist()[0]),
end=str(test.index.tolist()[-1]),
dynamic=True).apply(np.exp)), np.array(test),
args.loss)
# ordinary arima
loss_dict["arima"] = loss(
np.array(
ARIMA(train, (1, 0, 1)).fit(disp=0).predict(
start=str(test.index.tolist()[0]),
end=str(test.index.tolist()[-1]),
dynamic=True)), np.array(test), args.loss)
logger.info(loss_dict)
plt.figure(figsize=(12, 6))
loss_df = pd.DataFrame.from_dict(loss_dict, orient='index')
plt.bar(loss_df.index.tolist(), loss_df.iloc[:, 0])
plt.ylabel("RMSE")
plt.legend('')
plt.title("RMSE for Difference Models on Test Data")
plt.savefig("RMSE_model_comparison.png")
plt.close()
# prediction
logger.info("--------------prediction---------------")
prediction = model_predict(trend_model_fit, residual_model_fit, trend,
residual, seasonal, trend_diff_counts,
residual_diff_counts, True,
"2020-12-07 00:00:00", "2020-12-11 00:00:00",
args.period)
if args.log:
prediction = np.exp(prediction)
logger.info("2020-12-07 predicted value: " + str(prediction[0]))
logger.info("2020-12-08 predicted value: " + str(prediction[1]))
logger.info("2020-12-09 predicted value: " + str(prediction[2]))
logger.info("2020-12-10 predicted value: " + str(prediction[3]))
logger.info("2020-12-11 predicted value: " + str(prediction[4]))
logger.info("--------------Process ends-------------")
color='gray') # lowwer置信区间
plt.axhline(y=1.96 / np.sqrt(len(ts_log_diff)), linestyle='--',
color='gray') # upper置信区间
plt.title('Autocorrelation Function')
# p的获取:PACF图中曲线第一次穿过上置信区间.这里p取2
plt.subplot(122)
plt.plot(lag_pacf)
plt.axhline(y=0, linestyle='--', color='gray')
plt.axhline(y=-1.96 / np.sqrt(len(ts_log_diff)), linestyle='--', color='gray')
plt.axhline(y=1.96 / np.sqrt(len(ts_log_diff)), linestyle='--', color='gray')
plt.title('Partial Autocorrelation Function')
plt.tight_layout()
plt.show()
# AR model
model = ARIMA(ts_log, order=(2, 1, 0))
result_AR = model.fit(disp=-1)
plt.plot(ts_log_diff)
plt.plot(result_AR.fittedvalues, color='red')
plt.title('AR model RSS:%.4f' % sum(result_AR.fittedvalues - ts_log_diff)**2)
plt.show()
# MA model
model = ARIMA(ts_log, order=(0, 1, 2))
result_MA = model.fit(disp=-1)
plt.plot(ts_log_diff)
plt.plot(result_MA.fittedvalues, color='red')
plt.title('MA model RSS:%.4f' % sum(result_MA.fittedvalues - ts_log_diff)**2)
plt.show()
# ARIMA 将两个结合起来 效果更好
def arima_models(ts_log, p, d, q):
model = ARIMA(ts_log, order=(p, d, q))
results = model.fit(disp=-1)
return results
def TrainTimeSeries(dataset, p, d, q, freq):
if freq > 0:
model = ARIMA(dataset,order= (p,d,q))
return model
def gen_ohlcv(interval: int) -> go.Figure:
"""Generate OHLCV Chart for BTCUSD with predicted price overlay.
Args:
interval: update the graph based on an interval
"""
# hack to wrap interval around available data. OOS starts at 1500,
# df has a total of 2274 rows after processing to wrap around
# 2274-1500 ~ 750. Reset prediction data to empty df.
# interval = interval % 750
# print("interva is {}...".format(interval))
# read data from source
# df = get_ohlcv_data(interval - 100, interval)
df = bitfinex_candles_api()
df["log_ret"] = np.log(df.Close) - np.log(df.Close.shift(1))
print("\ndata df loaded, starting prediction...\n")
# online training and forecast.
model = ARIMA(df.tail(60)["log_ret"], order=(3, 1, 0)).fit(disp=0)
pred = model.forecast()[0]
print("\nprediction ended, writing to output df...")
# save forecast to output dataframe. should be dB irl.
next_dt = df.tail(1).index[0] + pd.Timedelta("1 minute")
config.df_pred.loc[next_dt] = [
pred[0],
(np.exp(pred) * df.tail(1).Close.values)[0],
]
print("\nnext datetime is {}...".format(next_dt))
# get index location of period.
loc = config.df_pred.index.get_loc(next_dt) + 1
print("\nloc is {}...".format(loc))
# slices for the past N periods perdiction for plotting
df_pred_plot = config.df_pred.iloc[slice(max(0, loc - 30),
min(loc,
len(df)))].sort_index()
print("\n set pred df for plotting...\n", df_pred_plot)
# plotting ohlc candlestick
trace_ohlc = go.Candlestick(
x=df.tail(50).index,
open=df["Open"].tail(50),
close=df["Close"].tail(50),
high=df["High"].tail(50),
low=df["Low"].tail(50),
opacity=0.5,
hoverinfo="skip",
name="BTCUSD",
)
# plotting prediction line
trace_line = go.Scatter(
x=df_pred_plot.index,
y=df_pred_plot.pred_Close,
line_color="yellow",
mode="lines+markers",
name="Predicted Close",
)
layout = go.Layout(
plot_bgcolor=config.app_color["graph_bg"],
paper_bgcolor=config.app_color["graph_bg"],
font={"color": "#fff"},
height=700,
xaxis={
"showline": False,
"showgrid": False,
"zeroline": False
},
yaxis={
"showgrid": True,
"showline": True,
"fixedrange": True,
"zeroline": True,
"gridcolor": config.app_color["graph_line"],
"title": "Price (USD$)",
},
)
return go.Figure(data=[trace_ohlc, trace_line], layout=layout)
def StartARIMAForecasting(Actual, P, D, Q):
# print('from function screaming')
model = ARIMA(Actual, order=(P, D, Q))
model_fit = model.fit(disp=0)
prediction = model_fit.forecast()[0]
return prediction
ptime.append(temp)
sumtime.append(len(temp))
#dataset
from statsmodels.tsa.arima_model import ARIMA
time = [float(i) for i in time]
time = pd.Series(time, index=tstamp)
#separate training-test set split
size = int(len(time) - 100)
train, test = time[0:size], time[size:len(time)]
history = [x for x in train]
predictions = list()
#train ARIMA model
model = ARIMA(history, order=(3, 1, 5))
model_fit = model.fit(disp=0)
#forecast the next 100 pieces of data
output = model_fit.forecast(steps=100)[0]
output = [x for x in output]
test = [x for x in test]
fig = plt.figure(figsize=(10, 5))
#plot the predicted vs. expected graph
ax = fig.add_subplot(111)
ax.plot(test, label="Observed")
ax.plot(output, label="Predicted")
plt.xlabel("Time")
plt.ylabel("Number of Crime")
data.append(train_x2)
data.append(train_x3)
#print(data[0])
#print(train_x)
for currentdata in data:
TS = currentdata
final_aic = math.inf
final_bic = math.inf
final_order = (0, 0, 0)
#print(final_order)
for p in range(0, 3):
for d in range(1, 3):
for q in range(0, 3):
try:
model = ARIMA(TS, order=(p, d, q))
#print(p,q,d)
results_ARIMA = model.fit(disp=-1)
current_aic = results_ARIMA.aic #compute AIC error on the model formed so far
current_bic = results_ARIMA.bic #compute BIC error on the model formed so far
#print(p,d,q)
if (
current_bic < final_bic and current_aic < final_aic
): #if current error is minimum then update all the order,model etc
final_aic = current_aic
final_bic = current_bic
final_order = (p, d, q)
'''results_final_ARIMA = final_arima.fit()
print(results_final_ARIMA.summary())
#final_accuracy = accuracy(model)'''
except (ValueError, RuntimeError, TypeError, NameError):
df.diff().plot() # In other words, we make the time serie "stationary" # ======================= ARIMA ======================= # ARIMA model is the combination of these two concepts. # ARIMA uses the correlation with previous time steps to make forecast. ## ARIMA = AutoRregression + I (remove trend) + Moving Average # This is the same concept as we have seen in visualization. # (Moving average is for errors, which we won't use here) from statsmodels.tsa.arima_model import ARIMA arima = ARIMA(df.consumption, order=(5,1,0)) model_fit = arima.fit() #(disp=0) prediction = model_fit.forecast() df.plot() prediction.plot() print(model_fit.summary()) # ================== Train/Test split ================== ''' END OF THE INTERMEDIATE COURSE '''
# rolling_mean = df_log.rolling(window=12).mean()
# df_log_minus_mean = df_log - rolling_mean
# df_log_minus_mean.dropna(inplace=True)
# get_stationarity(df_log_minus_mean)
#
# rolling_mean_exp_decay = df_log.ewm(halflife=12, min_periods=0, adjust=True).mean()
# df_log_exp_decay = df_log - rolling_mean_exp_decay
# df_log_exp_decay.dropna(inplace=True)
# get_stationarity(df_log_exp_decay)
#
# df_log_shift = df_log - df_log.shift()
# df_log_shift.dropna(inplace=True)
# get_stationarity(df_log_shift)
decomposition = seasonal_decompose(df_log)
model = ARIMA(df_log, order=(5, 1, 0))
results = model.fit(disp=-1)
#plt.plot(df_log_shift)
plt.plot(results.fittedvalues, color='red')
plt.show()
predictions_ARIMA_diff = pd.Series(results.fittedvalues, copy=True)
predictions_ARIMA_diff_cumsum = predictions_ARIMA_diff.cumsum()
predictions_ARIMA_log = pd.Series(df_log['PKB'].iloc[0], index=df_log.index)
predictions_ARIMA_log = predictions_ARIMA_log.add(
predictions_ARIMA_diff_cumsum, fill_value=0)
predictions_ARIMA = np.exp(predictions_ARIMA_log)
plt.plot(df)
predictions_ARIMA.head()
plt.plot(predictions_ARIMA)
from pandas.tools.plotting import autocorrelation_plot
def parser(x):
return datetime.strptime('190'+x, '%Y-%m')
series = read_csv('shampoo-sales.csv', header=0, parse_dates=[0], index_col=0, squeeze=True, date_parser=parser)
print(series.head())
series.plot()
pyplot.show()
autocorrelation_plot(series)
pyplot.show()
# fit model
model = ARIMA(series, order=(5,1,0))
model_fit = model.fit(disp=0)
print(model_fit.summary())
# plot residual errors
residuals = DataFrame(model_fit.resid)
residuals.plot()
pyplot.show()
residuals.plot(kind='kde')
pyplot.show()
print(residuals.describe())
# http://www.statsmodels.org/devel/generated/statsmodels.tsa.arima_model.ARIMA.predict.html
X = series.values
def time_series_analysis():
from pandas import datetime
def parser(x):
return datetime.strptime('190' + x, '%Y-%m')
# return datetime.strptime(x,'%Y-%m-%d')
index1 = randint(1, 1450) # hard coded
index2 = randint(1, 1412)
data, label = get_time_series_index_based_method1(index1)
series = pd.read_csv('elec.csv',
header=0,
parse_dates=[0],
index_col=0,
squeeze=True)
# series = { # 'time': pd.Series(['2011-01-01', '2011-01-02', '2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06']),
# 'time': pd.Series(['1901-01', '1902-01', '1903-01', '1904-01', '1905-01', '1906-01', '1907-01', '1908-01']),
# 'value': pd.Series([19330.143540669856600, 30641.148325358849700, 23813.397129186604700, 23272.727272727275100,
# 22866.028708133973200, 23961.722488038278900, 25856.459330143542400, 29598.086124401913600])}
# series = {
# 'time':pd.Series(data),
# 'value':pd.Series(label)
# }
# series = pd.DataFrame(series)
# series.plot()
# pyplot.show()
from sklearn.metrics import mean_squared_error
from pandas import read_csv
from statsmodels.tsa.arima_model import ARIMA
series1 = read_csv('shampoo.csv',
header=0,
parse_dates=[0],
index_col=0,
squeeze=True)
# series = { # 'time': pd.Series(['2011-01-01', '2011-01-02', '2011-01-03', '2011-01-04', '2011-01-05', '2011-01-06']),
# 'time': pd.Series(['1901-01', '1902-01', '1903-01', '1904-01', '1905-01', '1906-01', '1907-01', '1908-01']),
# 'value': pd.Series([19330.143540669856600, 30641.148325358849700, 23813.397129186604700, 23272.727272727275100,
# 22866.028708133973200, 23961.722488038278900, 25856.459330143542400, 29598.086124401913600])}
# fit model
# series = pd.DataFrame(series)
print("series1.head")
print(series1.head())
print("series.head")
print(series.head())
# series2 = read_csv('f**k.csv', header=0, parse_dates=[
# 0], index_col=0,squeeze=True)#\names=["day","value"])
series2 = read_csv('ElectricityBy15Minutes.csv',
header=0,
parse_dates=[0],
index_col=0,
squeeze=True,
names=["day", "value"])
model = ARIMA(series2, order=(6, 1, 0))
model_fit = model.fit(disp=0)
print(model_fit.summary())
# plot residual errors
residuals = DataFrame(model_fit.resid)
# residuals.plot()
# pyplot.show()
residuals.plot(kind='kde')
pyplot.show()
print(residuals.describe())
X = series2.values
print("X is ")
print(X)
size = int(len(X) * 0.66)
train, test = X[0:size], X[size:len(X)]
history = [x for x in train]
predictions = list()
import time
time.sleep(5)
for t in range(300):
model = ARIMA(history, order=(6, 1, 0))
model_fit = model.fit(disp=0)
output = model_fit.forecast()
yhat = output[0]
predictions.append(yhat)
obs = test[t]
history.append(obs)
print('predicted=%f, expected=%f' % (yhat, obs))
print("test-----")
print(test[:10])
print("history-----")
print(history[-10:])
print("predictions-----")
print(predictions)
error = mean_squared_error(history[-300:], predictions)
print('Test MSE: %.3f' % error)
print(predictions)
generate_comparison_plot(history[-300:], predictions)
print(cal_error(history[-300:], predictions))
#print(sc_rolmean,sc_rolstd)
check_adfuller(turb_ma_diff['turb(uS)'])
check_mean_std(turb_ma_diff, 'Turb(FNU)')
#%%
#X = series.values
train, test = turb_train.values, turb_test.values
turb_history = [x for x in train]
turb_predictions = list()
turb_diff = list()
k = 1921
for t in range(len(test)):
model = ARIMA(turb_history, order=(1, 0, 1))
model_fit = model.fit(disp=0)
output = model_fit.forecast()
yhat = output[0]
turb_predictions.append(yhat)
obs = test[t]
turb_history.append(obs)
diff = obs - yhat
turb_diff.append(diff)
print(
'TurbParameter Index= %d, predicted=%f, expected=%f, difference = %f' %
(k, yhat, obs, diff))
k = k + 1
if (k == 3000):
break
#test1, test2 = tts(test,test_size = 337, random_state=0, shuffle=False)
# load data
data = pd.read_csv('wholedata.csv')
series = data['value']
# prepare data
X = series.values
X = X.astype('float32')
train_size = int(len(X) * 0.99)
train, test = X[0:train_size], X[train_size:]
# walk-forward validation
history = [x for x in train]
predictions = list()
for i in range(len(test)):
# predict
model = ARIMA(history, order=(2,1,3))
model_fit = model.fit(trend='nc', disp=0)
yhat = model_fit.forecast()[0]
predictions.append(yhat)
# observation
obs = test[i]
history.append(obs)
# errors
residuals = [test[i]-predictions[i] for i in range(len(test))]
residuals = DataFrame(residuals)
print(residuals.describe())
pyplot.figure()
pyplot.subplot(211)
residuals.hist(ax=pyplot.gca())
pyplot.subplot(212)
residuals.plot(kind='kde', ax=pyplot.gca())
#
# Statsmodels also includes things like ARMA and ARIMA models that can be used to make predictions from time series. This data is not necessarily very stationary and often has strong periodic effects, so these may not necessarily work very well. I'll look at ARIMA predictions for the same set of very high viewcount pages.
# In[54]:
from statsmodels.tsa.arima_model import ARIMA
import warnings
cols = train.columns[1:-1]
for key in top_pages:
data = np.array(train.loc[top_pages[key], cols], 'f')
result = None
with warnings.catch_warnings():
warnings.filterwarnings('ignore')
try:
arima = ARIMA(data, [2, 1, 4])
result = arima.fit(disp=False)
except:
try:
arima = ARIMA(data, [2, 1, 2])
result = arima.fit(disp=False)
except:
print(train.loc[top_pages[key], 'Page'])
print('\tARIMA failed')
#print(result.params)
pred = result.predict(2, 599, typ='levels')
x = [i for i in range(600)]
i = 0
plt.plot(x[2:len(data)], data[2:], label='Data')
plt.plot(x[2:], pred, label='ARIMA Model')
from statsmodels.tsa.arima_model import ARIMA
import streamlit as st
import pandas as pd
import matplotlib.pyplot as plt
st.write("""
AdventureWorks Forecasting
""")
df = pd.read_csv('csv_data/forecast_data.csv', index_col=False)
df['RateChangeDate'] = pd.to_datetime(df['RateChangeDate'])
df.set_index('RateChangeDate', inplace=True)
ts = df['paidAmountSum']
st.text('This is a forecast')
st.line_chart(df)
st.dataframe(df)
model = ARIMA(ts, order=(1, 1, 1))
results = model.fit()
# results.plot_predict(1, 220)
values = st.sidebar.slider("Forecast Range", 200, 300)
st.pyplot(results.plot_predict(1, values))
#import the dependencies from random import random from statsmodels.tsa.arima_model import ARIMA #Generate randomized dataset in the range of 1 to 1000 dataset = [x + random() for x in range(1, 1000)] # # fitting the model arima = ARIMA(dataset, order=(1, 1, 1)) arima_fit = arima.fit(disp=False) # make prediction y = arima_fit.predict(len(dataset), len(dataset), typ='levels') print(y) #try to fiddle with the parameters of the ARIMA model
elif i == len(dataframe):
break
else:
i += 1
#remove_points_with_propagation(df["9"][:1000])
print('Counter: ' + str(counter))
print('Number of rows with 0 as value: ' +
str((df["9"] == 0).astype(int).sum(axis=0)))
series.plot()
plt.show()
#Creating model
model = ARIMA(series, order=(1, 0, 3))
model_fit = model.fit(disp=0)
print(model_fit.summary())
# Plot residual errors
residuals = pd.DataFrame(model_fit.resid)
fig, ax = plt.subplots(1, 2)
residuals.plot(title="Residuals", ax=ax[0])
residuals.plot(kind='kde', title='Density', ax=ax[1])
plt.show()
model_fit.plot_predict(dynamic=False)
plt.xlim(["2018-10-08", "2018-10-12"])
plt.ylabel("Voltage")
plt.show()
print('\t{}: {}'.format(key, value))
korona_log = np.log(korona_death)
plt.plot(korona_log)
plt.show()
rolling_mean = korona_log.rolling(window=12).mean()
korona_log_minus_mean = korona_log - rolling_mean
korona_log_minus_mean.dropna(inplace=True)
plt.plot(korona_log_minus_mean)
plt.show()
korona_log_shift = korona_log - korona_log.shift()
korona_log_shift.dropna(inplace=True)
plt.plot(korona_log_shift)
plt.show()
model = ARIMA(new_korona.death, order=(1, 1, 2))
model_fit = model.fit(disp=0)
print(model_fit.summary())
residuals = DataFrame(model_fit.resid)
residuals.plot()
pyplot.show()
residuals.plot(kind='kde')
pyplot.show()
print(residuals.describe())
model_fit.plot_predict(dynamic=False)
plt.show()
get_ipython().magic('matplotlib inline')
dta = (series - series.mean()) / (series.std())
plt.acorr(dta,maxlags = len(dta) -1, linestyle = "solid", usevlines = False, marker='')
plt.show()
autocorrelation_plot(series)
plt.show()
# In[29]:
from pandas import datetime
from pandas import DataFrame
from statsmodels.tsa.arima_model import ARIMA
# fit the model
arima_mod = ARIMA(series,order=(5,1,0))
arima_mod_fit = arima_mod.fit(disp=0)
print(arima_mod_fit.summary())
#Residual Errors
residuals = DataFrame(arima_mod_fit.resid)
residuals.plot()
plt.show()
residuals.plot(kind='kde')
plt.show()
print(residuals.describe())
# In[30]:
from pandas import read_csv
def arimaForecast(ts):
#X=ts['Close'].values
X = ts['high'].values
#size=int(len(X)*0.98)
size = len(X) - 100
train, test = X[0:size], X[size:len(X)]
test_length = len(test)
preds = []
history = [x for x in train]
backtest = Backtester(500, ts['high'].values, ts.index)
backtest.buy(train[-1], len(train) - 1)
bought_at = train[-1]
sold_at = None
i = 0
#print(history)
#print(test)
minGlobalDifference = math.inf
dailyforecast = []
globaldiffs = []
n_forecast = 0
while i < len(test):
print(i, " : ", test_length)
if i != 0 and i % 24 == 0:
plt.delaxes()
plt.bar(np.arange(len(dailyforecast)), dailyforecast)
plt.savefig(
"testingarimaplots/error_histogram{}.png".format(n_forecast))
print("SUMMARY OF PERIOD MEAN OF DIFFERENCES: {}".format(
mean(dailyforecast)))
if min(dailyforecast) < minGlobalDifference:
minGlobalDifference = min(dailyforecast)
print("STARTING FORECASTING PERIOD (DAY AHEAD)")
history = history[0:-24] + [x for x in test[i - 24:i]]
dailyforecast = []
n_forecast += 1
model = ARIMA(history, order=(3, 1, 4))
fit = model.fit(disp=0)
out = fit.forecast()
if out[0] >= bought_at * 1.18:
backtest.sell(test[i], len(train) + i)
sold_at = out[0]
if sold_at != None:
if out[0] <= sold_at * 0.82:
backtest.buy(test[i], len(train) + i)
bought_at = out[0]
if i % 100 == 0:
print("WEALTH {} ".format(backtest.getWealth(len(train) + i)))
if backtest.getWealth(len(train) + i) <= 53:
return
preds.append(out[0])
diff = abs(preds[-1] - test[i])
dailyforecast.append(diff[0])
globaldiffs.append(diff[0])
#if diff>=150: history.append(test[i])
#else: history.append(preds[-1])
history.append(test[i])
print("PREDICTED {} EXPECTED {} DIFFERENCE {}".format(
out[0], test[i], diff[0]))
i += 1
plt.plot(preds)
plt.plot(test)
plt.show()
print(
"RMSE: {}, SMALLEST DIFF BETWEEN REAL AND PREDICTED {} MEAN OF ABSOLUTE DIFFERENCES {}"
.format(mean_squared_error(test, preds), minGlobalDifference,
mean(globaldiffs)))
ts_log_diff.plot(figsize=(15, 6))
test_stationarity(ts_log_diff.dropna()) #.dropna(inplace=True))
# Using decomposition method to decompose time series
from pylab import rcParams
plt.figure(5)
rcParams['figure.figsize'] = 15, 6
#decomposition = sm.tsa.seasonal_decompose(ts_log, model = 'additive')
decomposition = sm.tsa.seasonal_decompose(ts_log, freq=12, model='additive')
decomposition.plot()
### Commented: ở đây decomposition ko có giá trị trả về. Tức là hàm seasonal_decompose ko trả ra kết quả gì cả
### nguyên nhân là do: gọi chuỗi dừng trong Arima, vì ts_log ko có tính dừng, nên khi gọi hàm này ra, kết quả ko có
# Build ARIMA model
arima_model = ARIMA(ts_log, order=(2, 1, 2))
arima_model_fit = arima_model.fit(disp=-1)
plt.figure(6)
plt.plot(ts_log_diff)
plt.plot(arima_model_fit.fittedvalues, color='red')
plt.title('RSS: %.4f' % np.nansum(
(arima_model_fit.fittedvalues - ts_log_diff)**2))
# Read summary of ARIMA model
print(arima_model_fit.summary())
# Convert predicted values to original scale
predictions_ARIMA_diff = pd.Series(arima_model_fit.fittedvalues, copy=True)
print(predictions_ARIMA_diff.head()
) # these are fitted values on the transformed data
plt.axhline(y=0,linestyle='--',color='gray')
#plot acf
plt.subplot(122)
plt.plot(lag_pacf)
plt.axhline(y=0,linestyle='--',color='gray')
# In[26]:
from statsmodels.tsa.arima_model import ARIMA
#AR Model
model = ARIMA(indexedDataset_logScale,order=(2, 1, 2))
# print(model)
result_AR= model.fit(disp=-1)
plt.plot(datasetLogDiffShifting)
plt.plot(result_AR.fittedvalues,color='red')
plt.title('RSS: %.4f'%sum((result_AR.fittedvalues-datasetLogDiffShifting["#Passengers"])**2))
print('Plotting AR Model')
# In[27]:
from statsmodels.tsa.arima_model import ARIMA
#MA Model
def ARIMA_model(time_series_diff, args, name):
"""
time_series_diff: stationary time_series after diff.
args: arguments parsed before.
name: the name of time_series_diff.
return fitted ARIMA model, parameters for ARIMA model.
"""
if args.plot and name in ["trend_diff", "residual_diff"]:
fig, axes = plt.subplots(1, 2, figsize=(16, 3), dpi=100)
plot_acf(time_series_diff.tolist(),
lags=min(50,
len(time_series_diff) - 1),
ax=axes[0])
plot_pacf(time_series_diff.tolist(),
lags=min(50,
len(time_series_diff) - 1),
ax=axes[1])
plt.savefig(name + "_acf_pacf.png")
plt.close()
# check if args.ic is illegal.
if args.ic not in ["bic", "aic"]:
logger.warning(
"The information criteria is illegal. Turn to default ic: BIC")
args.ic = "bic"
# check the value of convergence tol.
if args.tol > 0.01:
logger.warning(
"The convergence tolerance is too large. Turn to use default value: 1e-8"
)
args.tol = 1e-8
# check the likelihood function used.
if args.method not in ["css-mle", "mle", "css"]:
logger.warning(
"The likelihood function is illegal. Turn to default choice: css-mle"
)
args.method = "css-mle"
evaluate = sm.tsa.arma_order_select_ic(time_series_diff,
ic=args.ic,
trend="c",
max_ar=args.max_ar,
max_ma=args.max_ma)
# get the parameter for ARIMA model.
min_order = evaluate[args.ic + "_min_order"]
# initial the success_flag to false
success_flag = False
while not success_flag:
# construct the ARIMA model.
model = ARIMA(time_series_diff, order=(
min_order[0], 0, min_order[1]
)) # d is the order of diff, which we have done that perviously.
# keep finding initial parameters until convergence.
try:
model_fit = model.fit(
disp=False,
start_params=np.random.rand(min_order[0] + min_order[1] + 1),
method=args.method,
trend=
"c", # Some posts' experimentation suggests that ARIMA models may be less likely to converge with the trend term disabled, especially when using more than zero MA terms.
transparams=True,
solver=
"lbfgs", # we turn to use this one, which gives the best RMSE & executation time.
tol=args.tol, # The convergence tolerance. Default is 1e-08.
)
success_flag = True
except:
logger.warning("Error occurs, try another starting parameters.")
pass
return model_fit, min_order
plt.show()
from collections import deque
items = deque(np.asarray(fft_df['absolute'].tolist()))
items.rotate(int(np.floor(len(fft_df)/2)))
plt.figure(figsize=(10, 7), dpi=80)
plt.stem(items)
plt.title('Figure 4: Components of Fourier transforms')
plt.show()
from statsmodels.tsa.arima_model import ARIMA
from pandas import DataFrame
from pandas import datetime
series = data_FT['GS']
model = ARIMA(series, order=(5, 1, 0))
model_fit = model.fit(disp=0)
print(model_fit.summary())
from pandas.tools.plotting import autocorrelation_plot
autocorrelation_plot(series)
plt.figure(figsize=(10, 7), dpi=80)
plt.show()
plt.figure(figsize=(12, 6), dpi=100)
plt.plot(test, label='Real')
plt.plot(predictions, color='red', label='Predicted')
plt.xlabel('Days')
plt.ylabel('USD')
plt.title('Figure 5: ARIMA model on GS stock')
plt.legend()
plt.show()
def smape(y_true, y_pred):
return np.mean(
(np.abs(y_true - y_pred) * 200 / (np.abs(y_true) + np.abs(y_pred))))
# In[18]:
train_val = train_set['Open'].values
test_val = test_set['Open'].values
history = [x for x in train_val]
print(type(history)) #this is list of training data
prediction = list()
prediction
for t in range(len(test_val)):
model = ARIMA(history, order=(3, 1, 0))
model_fit = model.fit(disp=0)
output = model_fit.forecast()
yhat = output[0]
prediction.append(yhat)
obs = test_val[t]
history.append(obs)
error = mean_squared_error(test_val, prediction)
print("Mean squared error : %0.3f", error)
error2 = smape(test_val, prediction)
print("Symmetric mean absolute percentage error: %0.3f", error2)
# In[19]:
print('Testing Mean Squared Error: %.3f' % error)
print("Symmetric mean absolute percentage error: %0.3f" % error2)
#!/usr/bin/env python2 # -*- coding: utf-8 -*- from statsmodels.tsa.arima_model import ARIMA SERIES = [16, 20, 32, 40, 20, 18, 11, 21, 4, 6, 31, 48, 43, 49, 37] model = ARIMA(SERIES, order=(4, 1, 1)) model_fit = model.fit(disp=0) prediction = model_fit.predict(16, 19, typ='levels') print prediction
def parser(x):
return datetime.strptime('190' + x, '%Y-%m')
series = read_csv('shampoo-sales.csv',
header=0,
parse_dates=[0],
index_col=0,
squeeze=True,
date_parser=parser)
X = series.values
size = int(len(X) * 0.66)
train, test = X[0:size], X[size:len(X)]
history = [x for x in train]
predictions = list()
for t in range(len(test)):
model = ARIMA(history, order=(5, 1, 0))
model_fit = model.fit(disp=0)
output = model_fit.forecast()
yhat = output[0]
predictions.append(yhat)
obs = test[t]
history.append(obs)
print('predicted=%f, expected=%f' % (yhat, obs))
error = mean_squared_error(test, predictions)
print('Test MSE: %.3f' % error)
# plot
pyplot.plot(test)
pyplot.plot(predictions, color='red')
pyplot.show()
#-*- coding: utf-8 -*-
#确定最佳p、d、q值
import pandas as pd
#参数初始化
discfile = '../data/discdata_processed.xls'
data = pd.read_excel(discfile, index_col='COLLECTTIME')
data = data.iloc[:len(data) - 5] #不使用最后5个数据
xdata = data['CWXT_DB:184:D:\\']
from statsmodels.tsa.arima_model import ARIMA
#定阶
pmax = int(len(xdata) / 10) #一般阶数不超过length/10
qmax = int(len(xdata) / 10) #一般阶数不超过length/10
bic_matrix = [] #bic矩阵
for p in range(pmax + 1):
tmp = []
for q in range(qmax + 1):
try: #存在部分报错，所以用try来跳过报错。
tmp.append(ARIMA(xdata, (p, 1, q)).fit().bic)
except:
tmp.append(None)
bic_matrix.append(tmp)
bic_matrix = pd.DataFrame(bic_matrix) #从中可以找出最小值
p, q = bic_matrix.stack().idxmin() #先用stack展平，然后用idxmin找出最小值位置。
print(u'BIC最小的p值和q值为：%s、%s' % (p, q))
mpl.rcParams['font.sans-serif'] = [u'SimHei']
mpl.rcParams['axes.unicode_minus'] = False
x = data['Passengers'].astype(np.float)
x = np.log(x)
print x.head(10)
show = 'prime' # 'diff', 'ma', 'prime'
d = 1
diff = x - x.shift(periods=d)
ma = x.rolling(window=12).mean()
xma = x - ma
p = 2
q = 2
model = ARIMA(endog=x, order=(p, d, q)) # 自回归函数p,差分d,移动平均数q
arima = model.fit(disp=-1) # disp<0:不输出过程
prediction = arima.fittedvalues
print type(prediction)
y = prediction.cumsum() + x[0]
mse = ((x - y)**2).mean()
rmse = np.sqrt(mse)
plt.figure(facecolor='w')
if show == 'diff':
plt.plot(x, 'r-', lw=2, label=u'原始数据')
plt.plot(diff, 'g-', lw=2, label=u'%d阶差分' % d)
#plt.plot(prediction, 'r-', lw=2, label=u'预测数据')
title = u'乘客人数变化曲线 - 取对数'
elif show == 'ma':
#plt.plot(x, 'r-', lw=2, label=u'原始数据')
model_fit = model.fit(maxlag=1, method='mle', disp=-1) # make prediction yhat = model_fit.predict(0, len(response)+10) createPlot(yhat, response) # In[343]: # Autoregressive Integrated Moving Average # ## Autoregressive Integrated Moving Average # # http://www.statsmodels.org/dev/generated/statsmodels.tsa.arima_model.ARIMA.html#statsmodels.tsa.arima_model.ARIMA # # https://otexts.com/fpp2/non-seasonal-arima.html model = ARIMA(endog = response, order=(1, 0, 1)) model_fit = model.fit(disp=False) # make prediction yhat = model_fit.predict(0, len(response)+10) createPlot(yhat, response) # In[344]: # Seasonal Autoregressive Integrated Moving Average # ## Seasonal Autoregressive Integrated Moving Average # http://www.statsmodels.org/dev/generated/statsmodels.tsa.statespace.sarimax.SARIMAX.html # fit model model = SARIMAX(response, order=(1, 1, 1), seasonal_order=(1, 1, 1, 1))
