def plot_prevision(train, test ,p,d,q,P,D,Q, *args, **kwargs):
focused = kwargs.get('focused', None)
exog = kwargs.get('exog', None)
if focused is None:
focused="Weekly_Sales"
y = prepareDS(train, focused)
if exog is None:
ex = None
tex = None
else:
ex = prepareDS(train, exog)
tex = prepareDS(test, exog)
yhat = sarimax_fit(y,ex, p,d,q,P,D,Q)
prevision = yhat.get_forecast(steps=len(test.set_index('Date').resample('MS').mean()), exog=tex)
pred_ci = prevision.conf_int()
ax = y.plot(label='observed', figsize=(14, 7))
prevision.predicted_mean.plot(ax=ax, label='Forecast')
ax.fill_between(pred_ci.index,
pred_ci.iloc[:, 0],
pred_ci.iloc[:, 1], color='k', alpha=.25)
ax.set_xlabel('Date')
ax.set_ylabel(focused)
plt.legend()
plt.show()
def check_prevision(train, p,d,q,P,D,Q, *args, **kwargs):
focused = kwargs.get('focused', None)
exog = kwargs.get('exog', None)
if focused is None:
focused="Weekly_Sales"
y = prepareDS(train, focused)
if exog is None:
ex = None
else:
ex = prepareDS(train, exog)
yhat = sarimax_fit(y[:'2012-01-01'],ex[:'2012-01-01'], p,d,q,P,D,Q)
pred = yhat.get_forecast(steps=len(y['2012-01-01':]), exog=ex['2012-01-01':])
# pred = yhat.get_prediction(start=pd.to_datetime('2012-01-01'), dynamic=False)
prevision_ci = pred.conf_int()
ax = y.plot(label='observed')
pred.predicted_mean.plot(ax=ax, label='Predicted', alpha=.7)
ax.fill_between(prevision_ci.index,
prevision_ci.iloc[:, 0],
prevision_ci.iloc[:, 1], color='k', alpha=.2)
ax.set_xlabel('Date')
ax.set_ylabel(focused)
plt.legend()
plt.show()
y_forecasted = pred.predicted_mean
y_truth = y['2012-01-01':]
# Compute the mean square error
rms = sqrt(mean_squared_error(y_forecasted, y_truth))
print('Forecast Root Mean Squared Error {}'.format(round(rms, 2)))
return yhat
def showgraph(vdf, xaxis, yaxis, xlabel, ylabel, title):
vdf = vdf.set_index(xaxis)
for col in np.nditer(yaxis):
vdf[str(col)].plot()
plt.legend(loc=1)
plt.xlabel(xlabel)
plt.ylabel(ylabel)
plt.title(title)
plt.show()
def show_decompose(df, *args, **kwargs):
focused = kwargs.get('focused', None)
model = kwargs.get('model', None)
if focused is None: focused = "Weekly_Sales"
y = df[['Date', focused]].copy()
y = y.set_index('Date')
rcParams['figure.figsize'] = 18, 8
decomposition = tsa.seasonal_decompose(y, model=model)
fig = decomposition.plot()
plt.gcf().canvas.set_window_title(model.upper())
plt.show()
def corr_heatmap(df):
corr = df.corr()
# Generate a mask for the upper triangle
mask = np.zeros_like(corr, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True
# Set up the matplotlib figure
f, ax = plt.subplots(figsize=(11, 9))
# Generate a custom diverging colormap
cmap = sns.diverging_palette(220, 10, as_cmap=True)
# Draw the heatmap with the mask and correct aspect ratio
sns.heatmap(corr,
mask=mask,
cmap=cmap,
vmax=.3,
center=0,
square=True,
linewidths=.5,
cbar_kws={"shrink": .5})
plt.title('Correlation Heat Map')
plt.show()
def plotforecast(train, test):
traindt = train.set_index('Date')
testdt = test.set_index('Date')
testdt.fillna(value=000, inplace=True)
forecast_col='Weekly_Sales'
X = np.array(traindt.drop(columns=[forecast_col]))
X = scale(X)
newX = np.array(testdt)
newX = scale(newX)
y = np.array(traindt[forecast_col])
y = scale(y)
print(len(X), len(y))
clf = LinearRegression(n_jobs=-1)
clf.fit(X, y)
forecast_set = clf.predict(newX)
traindt['Forecast'] = np.nan
last_date = traindt.iloc[-1].name
last_unix = last_date.timestamp()
one_week = 86400*7
next_unix = last_unix + one_week
for i in forecast_set:
next_date = datetime.fromtimestamp(next_unix)
next_unix += 86400
traindt.loc[next_date] = [np.nan for _ in range(len(traindt.columns)-1)]+[i]
traindt[forecast_col].plot()
traindt['Forecast'].plot()
plt.legend(loc=4)
plt.xlabel('Date')
plt.ylabel('Revenue')
plt.legend(loc=1)
plt.show()