def test_get_actual_peaks():
actual_load_fname = 'ieso_ga_master_dataset_allWeather_updated2020.csv'
forecasts_fname = 'ga_forecasts_top_2.csv'
actual_load, forecasts = dataloader(actual_load_fname, forecasts_fname)
# Test get_actual_peaks
actual_peaks = get_actual_peaks(actual_load)
assert isinstance(actual_peaks, pd.DataFrame)
def test_summarize():
actual_load_fname = 'ieso_ga_master_dataset_allWeather_updated2020.csv'
forecasts_fname = 'ga_forecasts_top_2.csv'
actual_load, forecasts = dataloader(actual_load_fname, forecasts_fname)
# Test get_actual_peaks
actual_peaks = get_actual_peaks(actual_load)
summary = summarize(actual_peaks, forecasts, save_path='')
assert isinstance(summary, pd.DataFrame)
def test_summarize_top_n():
actual_load_fname = 'ieso_ga_master_dataset_allWeather_updated2020.csv'
forecasts_fname = 'ga_forecasts_top_2.csv'
actual_load, forecasts = dataloader(actual_load_fname, forecasts_fname)
# Test get_actual_peaks
actual_peaks = get_actual_peaks(actual_load)
top_n_results, top_n_sum = summarize_top_n(actual_peaks, forecasts, 1)
assert isinstance(top_n_results, pd.DataFrame)
assert isinstance(top_n_sum, pd.DataFrame) or isinstance(
top_n_sum, pd.Series)
def get_report(actual_load, forecasts, n_probs_to_use):
"""
Arguments:
actual_load:
A dataframe. The entire ground-truth demands.
forecasts:
A dataframe. The entire forecasting results.
n_probs_to_use:
Number of top probabilities to use on each day
Return:
A dataframe. The performance report.
This function generates the performance reports for top
1, top 5, top 10, and top 20 peak days.
"""
mapper = {}
attr_keys = [
'Season', 'Top_n_peaks', 'Top_n_probs', 'HitRate(%)', 'Performance(%)'
]
hour_keys = ['12', '15', '16', '17', '18', '19', '20']
keys = attr_keys + hour_keys
for key in keys:
mapper[key] = []
actual_peaks = get_actual_peaks(actual_load)
top_n_forecasts = get_top_n_forecasts(forecasts, n_probs_to_use)
for n_peaks_to_use in [1, 3, 5, 10, 20]:
top_n_results, _ = summarize_top_n(actual_peaks, forecasts,
n_peaks_to_use)
top_n_adjusted = adjust_probs(top_n_results, top_n_forecasts,
n_peaks_to_use, n_probs_to_use)
# Update mapper with each season
mapper = get_top_n_report(top_n_results, top_n_adjusted,
n_peaks_to_use, n_probs_to_use, hour_keys,
mapper)
report_df = pd.DataFrame(data=mapper)
return report_df
def test_adjust_probs():
actual_load_fname = 'ieso_ga_master_dataset_allWeather_updated2020.csv'
forecasts_fname = 'ga_forecasts_top_2.csv'
actual_load, forecasts = dataloader(actual_load_fname, forecasts_fname)
n_peaks_to_use = 1
n_probs_to_use = 3
# Get top n forecasts
top_n_forecasts = get_top_n_forecasts(forecasts, n_probs_to_use)
# Get actual peaks
actual_peaks = get_actual_peaks(actual_load)
# Get top n results
top_n_results, _ = summarize_top_n(actual_peaks, forecasts, n_peaks_to_use)
# Get top n adjusted probabilities
top_n_adjusted = adjust_probs(top_n_results, top_n_forecasts,
n_peaks_to_use, n_probs_to_use)
assert isinstance(top_n_adjusted, pd.DataFrame)
actual_load_fname = 'ieso_ga_master_dataset_allWeather_updated2020.csv'
forecasts_fname = 'ga_forecasts_top_2.csv'
actual_load, forecasts = dataloader(actual_load_fname, forecasts_fname)
n_peaks_to_use = 1
n_probs_to_use = 3
n_probs = forecasts_fname.split('_')[3].split('.')[0]
n_probs = int(n_probs)
if n_probs_to_use > n_probs:
print('n_probs_to_use must not exceed n_probs!')
n_probs_to_use = n_probs
# Get top n forecasts
top_n_forecasts = get_top_n_forecasts(forecasts, n_probs_to_use)
print('Top n forecasts:')
print(top_n_forecasts.head(3))
# Get actual peaks
actual_peaks = get_actual_peaks(actual_load)
# Get top n results
top_n_results, _ = summarize_top_n(actual_peaks, forecasts, n_peaks_to_use)
# Get top n adjusted probabilities
top_n_adjusted = adjust_probs(top_n_results, top_n_forecasts,
n_peaks_to_use, n_probs_to_use)
print('Top n adjusted:')
print(top_n_adjusted.head(3))