18] # same-day forecast, highlight, within [4,18]
highlight_X_1 = [
6, 7, 8, 9, 17, 18, 19
] # 1-day-ahead forecast, highlight the hour/month on the plot
else: # if DST[0]==False
highlight_X_0 = [7, 8, 9,
17] # same-day forecast, highlight, within [5,17]
highlight_X_1 = [6, 7, 8, 9, 17, 18, 19]
# ------------------#
show_plot = 1
save_plot = 0
# --------------------------------------------------------------------------------------------------------------------#
for k in k_list:
df = import_preprocess(station=k,
yr=year,
fulldata=True,
actual_diff=False,
fcst_diff=True)
# plot for a day
# dh = df[(df.daysahead == 1) & (df.month == 1) & (df.date == date(2019, 1, 2))]
# plt.plot(dh.ept, dh[y])
# for season_no, season in enumerate(seasons, 1): # start season index is 1
# By X boxplot
if by_X_boxplot:
sns.set_style("ticks")
for horizon in range(1, 2):
dh = df[(df.daysahead == horizon) & (df.DST.isin(DST))]
# if season_no == 1:
# --------------------------------#
check_actual_outlier_boxplot = 0 # check if there is any outliers/data issues in actual temperature data
check_daysahead_err_boxplot = 0 # check errors for each daysahead
# --------------------------------#
save_plot = 0
show_plot = 1
# zero_dif = 1 # find out when forecast Temp change is 0, what's the forecast error looks like
after_1000 = 0 # this is used for apple-to-apple comparison between same-day and 1-day-ahead. same-day only has forecasts after 1000am
highlight_month = []
# ----------------------------------------------------------------------------------------------------------------#
# %% Plot
for k in k_list:
if plot_scatter:
df = import_preprocess(station=k, yr=year)
for horizon in range(1, 2):
dh = df[df.daysahead == horizon]
if after_1000:
dh = df[(df.year.isin(year)) & (df.daysahead == horizon) & (df.hour > 10)]
# show average error on plots
avg_err = np.around(np.mean(df[(df.year.isin(year)) & (df.daysahead == horizon)].err), decimals=3)
# if zero_dif:
# for i in range(3): # col
# for j in range(3): # row
# colname = 'dif' + str(i + j * 3 + 1)
# zero = dh[dh[x + colname] == 0]
# freq = zero.groupby([x + colname, y]).size().reset_index(name="occurrance") # coordinate freq