test_predictor_sktime.py

import pandas as pd
import matplotlib.pyplot as plt

from sktime.datasets import load_airline
from sktime.forecasting.base import ForecastingHorizon
from sktime.forecasting.naive import NaiveForecaster
from sktime.forecasting.arima import AutoARIMA
from sktime.forecasting.ets import AutoETS

from sktime.performance_metrics.forecasting import MeanAbsolutePercentageError
from sktime.split import temporal_train_test_split
from sktime.split import ExpandingWindowSplitter
from sktime.forecasting.model_evaluation import evaluate
from sktime.utils.plotting import plot_series


from data_processing import read_datasets, add_production_field, interpolate_and_join, SolarParameters


parameters = SolarParameters()
met_2021_data, cons_2021_data = read_datasets()
add_production_field(met_2021_data, parameters)
all_data = interpolate_and_join(met_2021_data, cons_2021_data)

all_data['y'] = all_data['Consumption']
y = all_data[['y']]
y = y[y.index <= '2021-01-20']

print(len(y['y']), "data points read")

# 5 mins timestep means:
period = 12*24

forecaster = NaiveForecaster(strategy="last", sp=period)
# forecaster = AutoETS(auto=True, sp=period, n_jobs=-1)
# forecaster = AutoARIMA(sp=period, suppress_warnings=True)

cv = ExpandingWindowSplitter(
    step_length=period, fh=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], initial_window=period*10
    # step_length=period, fh=[1, 2], initial_window=period*2
)

strategy = "no-update_params"
df = evaluate(forecaster=forecaster, y=y, cv=cv, strategy=strategy, return_data=True)

print(df['y_pred'])

fig, ax = plot_series(
    y,
    df["y_pred"].iloc[0],
    df["y_pred"].iloc[1],
    df["y_pred"].iloc[2],
    df["y_pred"].iloc[3],
    df["y_pred"].iloc[4],
    df["y_pred"].iloc[5],
    markers=["o", "", "", "", "", "", ""],
    labels=["y_true"] + ["y_pred (Backtest " + str(x) + ")" for x in range(6)],
)
ax.legend()
plt.show()

exit()

y_train, y_test = temporal_train_test_split(y, test_size=len(y.index) // 2)

# step 2: running the basic forecasting workflow
fh = ForecastingHorizon(y_test.index, is_relative=False)



forecaster.fit(y_train)
y_pred = forecaster.predict(fh)

plot_series(y_train, y_test, y_pred, labels=["y_train", "y_test", "y_pred"])
plt.show()

# step 3: specifying the evaluation metric
mape = MeanAbsolutePercentageError(symmetric=False)
# if function interface is used, just use the function directly in step 4

# step 4: computing the forecast performance
print(mape(y_test, y_pred))