import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import statsmodels.api as sm
from tsdata import generate_sample_data
sample_ts, test_ts = generate_sample_data(trend=0.2, undiff=True)
ts_fig, ts_ax = plt.subplots(tight_layout=True)
sample_ts.plot(ax=ts_ax, c="b", label="Observed")
ts_ax.set_title("Training time series data")
ts_ax.set_xlabel("Date")
ts_ax.set_ylabel("Value")
diffs = sample_ts.diff().dropna()
ap_fig, (acf_ax, pacf_ax) = plt.subplots(2, 1, tight_layout=True, sharex=True)
sm.graphics.tsa.plot_acf(diffs, ax=acf_ax)
sm.graphics.tsa.plot_pacf(diffs, ax=pacf_ax)
acf_ax.set_ylabel("Value")
pacf_ax.set_xlabel("Lag")
pacf_ax.set_ylabel("Value")
import matplotlib.pyplot as plt
import statsmodels.api as sm
from tsdata import generate_sample_data
sample_ts, _ = generate_sample_data()
ts_fig, ts_ax = plt.subplots(tight_layout=True)
sample_ts.plot(ax=ts_ax, label="Observed")
ts_ax.set_title("Time series data")
ts_ax.set_xlabel("Date")
ts_ax.set_ylabel("Value")
adf_results = sm.tsa.adfuller(sample_ts)
adf_pvalue = adf_results[1]
print("Augmented Dickey-Fuller test:\nP-value:", adf_pvalue)
ap_fig, (acf_ax, pacf_ax) = plt.subplots(2, 1, sharex=True, tight_layout=True)
sm.graphics.tsa.plot_acf(sample_ts, ax=acf_ax, title="Observed autocorrelation")
sm.graphics.tsa.plot_pacf(sample_ts, ax=pacf_ax, title="Observed partial autocorrelation")
pacf_ax.set_xlabel("Lags")
pacf_ax.set_ylabel("Value")
acf_ax.set_ylabel("Value")
arma_model = sm.tsa.ARMA(sample_ts, order=(1, 1))
import pandas as pd
import matplotlib.pyplot as plt
from fbprophet import Prophet
from tsdata import generate_sample_data
sample_ts, test_ts = generate_sample_data(undiff=True, trend=0.2)
df_for_prophet = pd.DataFrame({
"ds": sample_ts.index, # dates
"y": sample_ts.values # values
})
model = Prophet()
model.fit(df_for_prophet)
forecast_df = model.make_future_dataframe(periods=50)
forecast = model.predict(forecast_df)
fig, ax = plt.subplots(tight_layout=True)
sample_ts.plot(ax=ax, label="Observed", title="Forecasts")
forecast.plot(x="ds", y="yhat", ax=ax, c="r", label="Predicted")
ax.fill_between(forecast["ds"].values,
forecast["yhat_lower"].values,
forecast["yhat_upper"].values,
color="r",
alpha=0.4)
test_ts.plot(ax=ax, c="k", label="Future")
ax.legend()
ax.set_xlabel("Date")
appear at regular intervals. This behavior is called seasonality in the
analysis of time series. It is relatively easy to adapt the standard ARIMA
model to incorporate seasonality, resulting in what is sometimes called a
SARIMA model.
This module illustrates how to model time series data that includes seasonal
behavior and use this model to produce forecasts.
"""
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import statsmodels.api as sm
from tsdata import generate_sample_data
sample_ts, test_ts = generate_sample_data(undiff=True, seasonal=True)
ts_fig, ts_ax = plt.subplots(tight_layout=True)
sample_ts.plot(ax=ts_ax, title="Time series", label="Observed")
ts_ax.set_xlabel("Date")
ts_ax.set_ylabel("Value")
ap_fig, (acf_ax, pacf_ax) = plt.subplots(2, 1, sharex=True, tight_layout=True)
sm.graphics.tsa.plot_acf(sample_ts, ax=acf_ax)
sm.graphics.tsa.plot_pacf(sample_ts, ax=pacf_ax)
pacf_ax.set_xlabel("Lag")
pacf_ax.set_ylabel("Value")
acf_ax.set_ylabel("Value")
diffs = sample_ts.diff().dropna()
dap_fig, (dacf_ax, dpacf_ax) = plt.subplots(2,