def qqplot(predicted, real, city, state, look_back, all_predict_n=False):
"""
Plot QQPlot for prediction values
:param predicted: Predicted matrix
:param real: Array of target_col values used in the prediction
:param city: Geocode of the target city predicted
:param state: State containing the city
:param look_back: Look-back time window length used by the model
:param all_predict_n: If True, plot the qqplot for every value predicted
:return:
"""
data_full = get_alerta_table(city, state)
data_full = data_full[['casos_est', 'SE']].reset_index(drop=True)
data_full.SE = [str(i)[-2:] for i in data_full.SE]
rvs = create_rvs(data_full)
if all_predict_n:
fig, axs = P.subplots(4, 1, figsize=[6, 20])
weeks = list(range(predicted.shape[1]))
else:
fig, axs = P.subplots(1, 1, figsize=[6, 5])
weeks = [0]
for week in weeks:
data = data_full.iloc[look_back + week:len(real) + look_back + week]
data['preds'] = predicted[:, week]
data['real'] = real[:, week]
data = data.merge(rvs, on='SE', how='left')
data['p_preds'] = data.apply(create_px, axis=1)
data['q_real'] = data.apply(create_qy, axis=1)
train = int(len(real) * 0.7)
preds_max = int(data['p_preds'].max())
axs.plot(list(range(preds_max)), list(range(preds_max)), color='black')
data[train:].plot(x='p_preds',
y='q_real',
kind='scatter',
ax=axs,
color='b',
legend=True)
data[:train].plot(x='p_preds',
y='q_real',
kind='scatter',
ax=axs,
color='r',
legend=True)
P.savefig('{}/qqplot_{}'.format(FIG_PATH, city), dpi=300)
# P.show()
return data
import numpy as np
import pandas as pd
import pyflux as pf
from datetime import datetime
import matplotlib.pyplot as plt
from infodenguepredict.data.infodengue import get_alerta_table
def build_model(data, ar=2, sc=4, family=pf.families.Poisson, target=None):
model = pf.GASLLEV(data=data, family=family(), target=target)
return model
if __name__ == "__main__":
prediction_window = 5 # weeks
data = get_alerta_table(3304557) # Nova Iguaçu: 3303609
data.casos.plot()
model = build_model(data, ar=2, sc=6, target='casos')
fit = model.fit() #'BBVI',iterations=1000,optimizer='RMSProp')
print(fit.summary())
model.plot_fit()
model.plot_z(figsize=(15, 5))
plt.savefig('GASLLEV_in_sample.png')
data.casos.plot(style='ko')
model.plot_predict(h=10, past_values=52)
plt.savefig('GASLLEV_prediction.png')
# plt.show()
def get_cities_from_state(state):
alerta_table = get_alerta_table(state=state)
cities_list = alerta_table.municipio_geocodigo.unique()
return cities_list
from infodenguepredict.data.infodengue import get_alerta_table, build_multicity_dataset
import seaborn
def build_model(data, lag_order, window_type):
data.index = pd.DatetimeIndex(data.index)
model = DynamicVAR(data, lag_order=lag_order, window=12, window_type=window_type)
return model
if __name__ == "__main__":
prediction_window = 6 # weeks
# scenario = 'local'
scenario = 'global'
if scenario == 'local':
data = get_alerta_table(3303500) # Nova Iguaçu: 3303500
data = data[['casos', 'p_inc100k','nivel']]
else:
data = build_multicity_dataset('RJ')
data = data[[col for col in data.columns if col.startswith('casos') and not col.startswith('casos_est')][:3]]
data = data.diff()
print(data.info())
#TODO: Apply Seasonal differencing to series
# data.casos.plot(title="Series")
model = build_model(data, 12, 'expanding')
# fit = model.(maxlags=11, ic='aic') # 4 lags
# print(model.coefs.minor_xs('casos_3303500').info())
forecast = model.forecast(prediction_window)
def test_alerta_state(self):
df = get_alerta_table(state='RJ')
self.assertIsInstance(df, pd.DataFrame)
self.assertGreater(df.size, 0)
assert (df.municipio_geocodigo.values > 3300000).all()
assert (df.municipio_geocodigo.values < 4000000).all()
def test_alerta_all_munic(self):
df = get_alerta_table()
self.assertIsInstance(df, pd.DataFrame)
self.assertGreater(df.size, 0)
self.assertGreater(df.municipio_geocodigo.value_counts().size, 1)
def test_alerta_one_munic(self):
df = get_alerta_table(3303609)
self.assertIsInstance(df, pd.DataFrame)
self.assertGreater(df.size, 0)
self.assertGreater(df.casos.sum(), 0)
self.assertEquals(df.municipio_geocodigo.value_counts().size, 1)
