plt.rc('axes', facecolor='#E6E6E6')
plt.rc('lines', linewidth=0.8)
ml_model_names = ['RandomForest','XGBoost','LogisticRegression']
time = 'second_hour'
target_set = ['tornado', 'severe_hail', 'severe_wind']
normalize_method = {'tornado': 'standard', 'severe_hail' : None, 'severe_wind' : None}
resample_method = {'tornado': 'under', 'severe_hail' : None, 'severe_wind' : None}
imputer_method = 'simple'
line_colors = plt_config.colors_for_ml_models_dict
line_labels = ml_model_names
fig_fname = f'performance_diagram.png'
#################
base_plot = PlotStructure()
lead_times = np.arange(12)*5 + 60
plt.rc("xtick", labelsize=7) # fontsize of the x-axis tick marks
plt.rc("ytick", labelsize=7) # fontsize of the y-axis tick marks
vars_to_load = ['AUC_mean', 'AUPDC_NORM_mean', 'CSI_mean', 'BSS_mean']
n_panels = 12
figsize = (10,5)
fig, axes = base_plot.create_subplots(
n_panels = n_panels,
figsize = figsize,
n_columns = 3,
sharex=True,
wspace=0.25,
var_set = ['POD', 'POFD']
mean_vars = [v + '_mean' for v in var_set]
lower_bounds = [v + '_lowerbound' for v in var_set]
upper_bounds = [v + '_upperbound' for v in var_set]
vars_to_load = mean_vars + lower_bounds + upper_bounds + [
'AUC_mean', 'predictions', 'targets'
]
line_colors = plt_config.colors_for_ml_models_dict
line_labels = ml_model_names
fig_fname = f'roc_diagram.png'
#################
base_plot = PlotStructure()
n_panels = len(times) * len(target_set)
if n_panels <= 3:
figsize = (12, 4)
else:
figsize = (10, 5)
fig, axes = base_plot.create_subplots(n_panels=n_panels,
figsize=figsize,
sharex=True,
sharey=True,
wspace=0.1,
hspace=0.1)
ax_iterator = base_plot.axes_to_iterator(n_panels, axes)
from base_plotting import PlotStructure
from wofs_ml.io.io import IO, target_vars, uh_baseline_probs, hail_baseline_probs, wnd_baseline_probs
from wofs_ml.io.cross_validation_generator import DateBasedCV
from wofs_ml.evaluate.metrics import brier_skill_score, bss_reliability
from sklearn.metrics import roc_auc_score, average_precision_score
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.ticker import (MaxNLocator, FormatStrFormatter,
AutoMinorLocator)
import os
os.environ['KMP_WARNINGS'] = '0'
base_plot = PlotStructure()
io = IO()
time_set = ['first_hour', 'second_hour']
times = ['FIRST HOUR', 'SECOND HOUR']
target_set = ['tornado', 'severe_hail', 'severe_wind']
thresholds = {'tornado': 180, 'severe_hail': 120, 'severe_wind': 80}
baseline_thresholds = {
'tornado': np.arange(50, 260, 10),
'severe_hail': np.arange(50, 260, 10),
'severe_wind': np.arange(50, 260, 10)
}
baseline_per_target = {
from base_plotting import PlotStructure
from pymint import InterpretToolkit
from wofs_ml.common.load_results import _load_train_data
from wofs.util.feature_names import to_readable_names, get_units
import matplotlib.pyplot as plt
from matplotlib.lines import Line2D
target = 'severe_wind'
time = 'first_hour'
mode = 'multipass'
metric = 'norm_aupdc'
perm_method = 'backward'
drop_opt = 'L1_based_feature_selection_with_manual'
n_vars = 12
base_plt = PlotStructure()
plt.rc("axes", labelsize=8)
def plot_ecdf(df, var, target, ax):
"""
Plot 1D ECDF
"""
sns.ecdfplot(
ax=ax,
data=df,
x=var,
hue=target,
legend=False,
)
from wofs_ml.io.io import IO
import os
from wofs.data.loadMRMSData import MRMSData
from wofs.data.loadEnsembleData import EnsembleData
from base_plotting import PlotStructure
import os
import xarray as xr
from wofs.processing.ObjectMatching import ObjectMatching, match_to_lsrs
from skimage.measure import regionprops
import pandas as pd
from skimage.color import label2rgb
from wofs.data.loadWWAs import loadWWA, is_severe, load_reports
base_plot = PlotStructure()
mrms = MRMSData()
get_time = personal_datetime()
io = IO()
date = '20180501'
time = '2330'
fcst_time_idx = 0
valid_date_and_time, initial_date_and_time = get_time.determine_forecast_valid_datetime(
date_dir=str(date), time_dir=time, fcst_time_idx=fcst_time_idx)
mrms_dbz = mrms.load_single_mrms_time(date_dir=date,
valid_datetime=initial_date_and_time,
from wofs.data.loadWWAs import loadWWA
import matplotlib.pyplot as matplt
import wofs.plotting.plotting_config as plt_config
from wofs.plotting.Plot import Plotting
import os
from wofs.data.loadMRMSData import MRMSData
from wofs.data.loadEnsembleData import EnsembleData
from base_plotting import PlotStructure
from load_predictions import load_ml_predictions, load_baseline, get_timestampes
get_time = personal_datetime()
mrms = MRMSData()
base_plot = PlotStructure()
target_cmap = {'tornado': Reds, 'severe_hail': Greens, 'severe_wind': Blues}
baseline_names = {
'tornado': 'Mid-level UH' + "\n" + r'> 180 m$^2$ s$^{-2}$',
'severe_hail': 'HAILCAST > 1 in',
'severe_wind': '80-m AGL Wind > 40 kts.'
}
prob_threshold = {
'tornado': 15.,
'severe_hail': 20.,
'severe_wind': 20.,
}
var_set = ['POD', 'SR']
mean_vars = [v+'_mean' for v in var_set]
lower_bounds = [v+'_lowerbound' for v in var_set]
upper_bounds = [v+'_upperbound' for v in var_set]
vars_to_load = mean_vars+lower_bounds+upper_bounds+['AUPDC_NORM_mean', 'targets', 'NCSI_mean', 'FREQ_BIAS_mean']
line_colors = plt_config.colors_for_ml_models_dict
line_labels = ml_model_names
fig_fname = f'performance_diagram_{drop_opt}.png'
add_ci=False
#################
base_plot = PlotStructure()
n_panels = len(times)*len(target_set)
if n_panels <= 3:
figsize = (12, 4)
else:
figsize = (10,5)
fig, axes = base_plot.create_subplots(
n_panels = n_panels,
figsize = figsize,
sharex=True,
sharey=True,
wspace=0.15,
hspace=0.15
)
from wofs.util.basic_functions import personal_datetime
from wofs.data.loadLSRs import loadLSR
from wofs_ml.common.load_results import _load_train_data, _load_test_data, _load_model, just_transforms
import matplotlib.pyplot as matplt
import wofs.plotting.plotting_config as plt_config
from wofs.plotting.Plot import Plotting
from wofs_ml.io.io import IO, base_vars_to_drop
import os
from wofs.data.loadMRMSData import MRMSData
from wofs.data.loadEnsembleData import EnsembleData
from base_plotting import PlotStructure
from copy import deepcopy
base_plot = PlotStructure()
mrms = MRMSData()
get_time = personal_datetime( )
io = IO()
model_name = 'RandomForest'
target_set = ['tornado']
colorbar_labels = ['Probability of Tornado']
target_cmap = {'tornado' : Reds, 'severe_hail' : Greens, 'severe_wind' : Blues}
time_data = 'first_hour'
normalize_method = {'tornado': 'standard', 'severe_hail' : None, 'severe_wind' : None}
resample_method = {'tornado': 'under', 'severe_hail' : None, 'severe_wind' : None}
vars_to_load = mean_vars + lower_bounds + upper_bounds + [
'AUPDC_NORM_mean',
'AUC_mean',
'predictions',
'targets',
'NCSI_mean',
'FREQ_BIAS_mean',
'BSS_mean',
]
line_colors = plt_config.colors_for_ml_models_dict
line_labels = ml_model_names
fig_fname = f'all_verification_diagram_{drop_opt}.png'
#################
base_plot = PlotStructure()
n_panels = 9
figsize = (9, 8)
fig, axes = base_plot.create_subplots(n_panels=n_panels,
figsize=figsize,
sharex=True,
sharey=True,
wspace=0.15,
hspace=0.2)
ax_iterator = base_plot.axes_to_iterator(n_panels, axes)
fontsize = 8
ax_iterator[0].set_ylabel('Probability of Detection', fontsize=fontsize)
ax_iterator[3].set_ylabel('Probability of Detection', fontsize=fontsize)
ax_iterator[6].set_ylabel('Conditional Event Frequency', fontsize=fontsize)
########################################
# USER-DEFINED PARAMETERS #
########################################
model_set = ['RandomForest', 'LogisticRegression', 'XGBoost']
time = 'first_hour'
pipeline_set = [
('severe_hail', None, None),
('severe_wind', None, None),
('tornado', 'under', 'standard'),
]
imputer_method = 'simple'
drop_high_corr_pred = True
#######################################
base_plot = PlotStructure()
n_panels = len(model_set) * len(pipeline_set)
if n_panels < 3:
figsize = (5, 4)
else:
figsize = (8, 5)
fig, axes = base_plot.create_subplots(n_panels=n_panels,
figsize=figsize,
sharex=True,
sharey=True,
wspace=0.15,
hspace=0.15)
ax_iterator = base_plot.axes_to_iterator(n_panels, axes)