def plot_2d_rmse_heatmap(self, cols):
#adjust for log params if necessary
xcol = cols[0]
ycol = cols[1]
xdata = np.array(self.flat_results[xcol])
xlabel = xcol
if self.is_log_param(xcol):
xdata_raw = np.array(self.flat_results[xcol].values, 'float')
xdata = np.log10(xdata_raw)
xlabel = "log10 %s" % xcol
ydata = np.array(self.flat_results[ycol])
ylabel = ycol
if self.is_log_param(ycol):
ydata_raw = np.array(self.flat_results[ycol].values, 'float')
ydata = np.log10(ydata_raw)
ylabel = "log10 %s" % ycol
kwargs = dict()
kwargs['xlabel'] = xlabel
kwargs['ylabel'] = ylabel
kwargs['labellist'] = ['xy', 'rmse']
kwargs['xdatalist'] = [xdata, xdata]
kwargs['ydatalist'] = [ydata, self.flat_results['rmse']]
kwargs['xerrlist'] = [None, None]
kwargs['yerrlist'] = [None, None]
kwargs['notelist'] = list()
kwargs['guideline'] = 0
plotlabel = "rmse_heatmap"
kwargs['plotlabel'] = plotlabel
kwargs['save_path'] = self.save_path
myph = PlotHelper(**kwargs)
myph.plot_2d_rmse_heatmap()
self.readme_list.append("Plot %s.png created\n" % plotlabel)
return
def plot(self):
self.readme_list.append("----- Plotting -----\n")
notelist = list()
notelist.append("Mean RMSE: {:.2f}".format(
self.statistics['avg_rmse']))
kwargs = dict()
kwargs['xlabel'] = self.xlabel
kwargs['ylabel'] = self.ylabel
kwargs['notelist'] = notelist
kwargs['save_path'] = self.save_path
kwargs['marklargest'] = self.mark_outlying_points
group_label_list = list()
rms_list = list()
group_rms_list = list()
for cvtest in self.cvtest_dict.keys():
group_rms_list.append((self.cvtest_dict[cvtest]['group'],
self.cvtest_dict[cvtest]['rmse']))
group_rms_list.sort() #sorts by group
group_rms_array = np.array(group_rms_list)
kwargs['xdatalist'] = [group_rms_array[:, 0]]
kwargs['ydatalist'] = [np.array(group_rms_array[:, 1], 'float')]
kwargs['xerrlist'] = [None]
kwargs['yerrlist'] = [None]
kwargs['labellist'] = ['predicted_rmse']
kwargs['plotlabel'] = "leave_out_group"
myph = PlotHelper(**kwargs)
myph.plot_rmse_vs_text()
self.readme_list.append("Plot leave_out_group.png created.\n")
return
def plot_single_rmse(self, col):
#adjust for log params if necessary
xdata = self.flat_results[col]
xlabel = col
if self.is_log_param(col):
import numpy as np
xdata_raw = np.array(self.flat_results[col].values, 'float')
xdata = np.log10(xdata_raw)
xlabel = "log10 %s" % col
kwargs = dict()
kwargs['xlabel'] = xlabel
kwargs['ylabel'] = 'RMSE'
kwargs['labellist'] = [xlabel]
kwargs['xdatalist'] = [xdata]
kwargs['ydatalist'] = [self.flat_results['rmse']]
kwargs['xerrlist'] = list([None])
kwargs['yerrlist'] = list([None])
kwargs['notelist'] = list()
kwargs['guideline'] = 0
plotlabel = "rmse_vs_%s" % col
plotlabel = plotlabel.replace(".", "_") #mask periods
kwargs['plotlabel'] = plotlabel
kwargs['save_path'] = self.save_path
myph = PlotHelper(**kwargs)
myph.multiple_overlay()
self.readme_list.append("Plot %s.png created\n" % plotlabel)
return
def plot_results(self, notelist=list()):
kwargs2 = dict()
kwargs2['xlabel'] = self.xlabel
kwargs2['ylabel'] = self.ylabel
kwargs2['labellist'] = list(["loo_prediction"])
kwargs2['xdatalist'] = list([self.testing_dataset.target_data])
kwargs2['ydatalist'] = list(
[self.cvtest_dict[0]['prediction_array']]
) #only one cvtest, with number of folds equal to number of data points
kwargs2['xerrlist'] = list([self.testing_dataset.target_error_data])
kwargs2['yerrlist'] = list([None])
kwargs2['notelist'] = list(notelist)
kwargs2['guideline'] = 1
kwargs2['plotlabel'] = "loo_results"
kwargs2['save_path'] = self.save_path
if not (self.mark_outlying_points is None):
kwargs2['marklargest'] = self.mark_outlying_points
if (self.testing_dataset.labeling_features is None):
raise ValueError(
"Must specify some labeling features if you want to mark the largest outlying points"
)
labels = self.testing_dataset.data[
self.testing_dataset.labeling_features[0]]
kwargs2['mlabellist'] = list([labels])
myph = PlotHelper(**kwargs2)
myph.multiple_overlay()
self.readme_list.append("Plot loo_results.png created,\n")
self.readme_list.append(" showing results of all LOO tests.\n")
return
def plot_meancv_overlay(self, notelist=list()):
kwargs2 = dict()
kwargs2['xlabel'] = self.xlabel
kwargs2['ylabel'] = self.ylabel
kwargs2['labellist'] = ["Mean CV test"]
kwargs2['xdatalist'] = list([self.testing_dataset.target_data])
kwargs2['ydatalist'] = list([self.statistics['average_prediction']])
kwargs2['xerrlist'] = list([None])
kwargs2['yerrlist'] = list([self.statistics['std_err']])
kwargs2['notelist'] = list(notelist)
kwargs2['guideline'] = 1
kwargs2['plotlabel'] = "mean_cv_overlay"
kwargs2['save_path'] = self.save_path
#kwargs2['std_err_in_mean'] = std_err_in_mean
if not (self.mark_outlying_points is None):
kwargs2['marklargest'] = self.mark_outlying_points
if self.testing_dataset.labeling_features is None:
raise ValueError(
"Must specify some labeling features if you want to mark the largest outlying points"
)
labels = self.testing_dataset.data[
self.testing_dataset.labeling_features[0]]
kwargs2['mlabellist'] = list([labels, labels])
myph = PlotHelper(**kwargs2)
myph.multiple_overlay()
self.readme_list.append("Plot mean_cv_overlay.png created,\n")
self.readme_list.append(" showing the mean cv of %i tests.\n" %
self.num_cvtests)
return
def plot_best_worst_overlay(self, notelist=list()):
kwargs2 = dict()
kwargs2['xlabel'] = self.xlabel
kwargs2['ylabel'] = self.ylabel
kwargs2['labellist'] = ["Best test", "Worst test"]
kwargs2['xdatalist'] = list([
self.testing_dataset.target_data, self.testing_dataset.target_data
])
kwargs2['ydatalist'] = list([
self.cvtest_dict[self.best_test_index]['prediction_array'],
self.cvtest_dict[self.worst_test_index]['prediction_array']
])
kwargs2['xerrlist'] = list([None, None])
kwargs2['yerrlist'] = list([None, None])
kwargs2['notelist'] = list(notelist)
kwargs2['guideline'] = 1
kwargs2['plotlabel'] = "best_worst_overlay"
kwargs2['save_path'] = self.save_path
if not (self.mark_outlying_points is None):
kwargs2['marklargest'] = self.mark_outlying_points
if self.testing_dataset.labeling_features is None:
raise ValueError(
"Must specify some labeling features if you want to mark the largest outlying points"
)
labels = self.testing_dataset.data[
self.testing_dataset.labeling_features[0]]
kwargs2['mlabellist'] = list([labels, labels])
myph = PlotHelper(**kwargs2)
myph.multiple_overlay()
self.readme_list.append("Plot best_worst_overlay.png created,\n")
self.readme_list.append(
" showing the best and worst of %i tests.\n" % self.num_cvtests)
return
def one_plot(self, group=None):
xdatalist = list()
xerrlist = list()
ydatalist = list()
yerrlist = list()
for testing_dataset in self.testing_datasets:
xdata = testing_dataset.data[self.feature_plot_feature]
ydata = testing_dataset.target_data
yerrdata = testing_dataset.target_error_data
xerrdata = None
if not (group is None):
xdata = xdata[testing_dataset.group_data == group]
ydata = ydata[testing_dataset.group_data == group]
if len(xdata) == 0:
xdata = list()
if len(ydata) == 0:
ydata = list()
if not (yerrdata is None):
yerrdata = yerrdata[testing_dataset.group_data == group]
if len(yerrdata) == 0:
yerrdata = None
xdatalist.append(xdata)
xerrlist.append(xerrdata)
ydatalist.append(ydata)
yerrlist.append(yerrdata)
kwargs = dict()
if group is None:
kwargs['plotlabel'] = "all"
if not (group is None):
kwargs['plotlabel'] = group
kwargs['xdatalist'] = xdatalist
kwargs['ydatalist'] = ydatalist
kwargs['xerrlist'] = xerrlist
kwargs['yerrlist'] = yerrlist
kwargs['labellist'] = self.data_labels
#kwargs['faces'] = ["None"] * len(self.data_labels)
kwargs['save_path'] = self.save_path
kwargs['xlabel'] = self.xlabel
kwargs['ylabel'] = self.ylabel
notelist = list()
#if not(self.plot_filter_out is None):
# notelist.append("Data not shown:")
# for (feature, symbol, threshold) in self.plot_filter_out:
# notelist.append(" %s %s %s" % (feature, symbol, threshold))
kwargs['notelist'] = notelist
myph = PlotHelper(**kwargs)
myph.multiple_overlay()
#if do_fft == 1:
# plotxy.single(xdata, fft(ydata), **kwargs)
return
def plot_group_splits_with_outliers(self,
group_dict=None,
outlying_groups=list(),
label="group_splits",
group_notelist=list()):
addl_kwargs = dict()
addl_kwargs['xlabel'] = self.xlabel
addl_kwargs['ylabel'] = self.ylabel
addl_kwargs['save_path'] = os.path.join(self.save_path, label)
addl_kwargs['guideline'] = 1
addl_kwargs['group_dict'] = group_dict
addl_kwargs['outlying_groups'] = list(outlying_groups)
addl_kwargs['plotlabel'] = label
addl_kwargs['notelist'] = list(group_notelist)
myph = PlotHelper(**addl_kwargs)
myph.plot_group_splits_with_outliers()
return
def plot_results(self, addl_plot_kwargs=None):
self.readme_list.append("----- Plotting -----\n")
if self.testing_dataset.target_data is None:
logger.warning(
"No testing target data. Predicted vs. measured plot will not be plotted."
)
self.readme_list.append("No target data.\n")
self.readme_list.append(
"No plot comparing predicted vs. measured data was made.\n")
return
plot_kwargs = dict()
plot_kwargs['xlabel'] = self.xlabel
plot_kwargs['ylabel'] = self.ylabel
plot_kwargs['plotlabel'] = "single_fit"
plot_kwargs['guideline'] = 1
notelist = list()
notelist.append("RMSE: %3.3f" % self.statistics['rmse'])
notelist.append("R-squared: %3.3f" % self.statistics['rsquared'])
notelist.append("R-squared (no int): %3.3f" %
self.statistics['rsquared_noint'])
notelist.append("Mean error: %3.3f" % self.statistics['mean_error'])
notelist.append("Mean abs error: %3.3f" %
self.statistics['mean_absolute_error'])
plot_kwargs['notelist'] = notelist
plot_kwargs['save_path'] = self.save_path
if not (addl_plot_kwargs is None):
for addl_plot_kwarg in addl_plot_kwargs:
plot_kwargs[addl_plot_kwarg] = addl_plot_kwargs[
addl_plot_kwarg]
if not (self.plot_filter_out is None):
self.readme_list.append("Plot filtering out:\n")
for (feature, symbol, threshold) in self.plot_filter_out:
self.readme_list.append(" %s %s %s\n" %
(feature, symbol, threshold))
notelist.append("Shown-only RMSE: %3.3f" %
self.statistics['filtered_rmse'])
notelist.append("Data not shown:")
for (feature, symbol, threshold) in self.plot_filter_out:
notelist.append(" %s %s %s" % (feature, symbol, threshold))
#Data should already have been filtered by now
plot_kwargs['xdatalist'] = [self.testing_dataset.target_data]
plot_kwargs['ydatalist'] = [self.testing_dataset.target_prediction]
if self.testing_dataset.target_error_feature is None:
plot_kwargs['xerrlist'] = [None]
else:
plot_kwargs['xerrlist'] = [self.testing_dataset.target_error_data]
if self.trained_model.__class__.__name__ == "GaussianProcessRegressor":
plot_kwargs['yerrlist'] = [
self.testing_dataset.target_prediction_sigma
]
else:
plot_kwargs['yerrlist'] = [None]
plot_kwargs['labellist'] = ["predicted_vs_measured"]
myph = PlotHelper(**plot_kwargs)
myph.multiple_overlay()
self.readme_list.append("Plot single_fit.png created.\n")
self.readme_list.append(
" Plotted data is in the data_... csv file.\n")
self.readme_list.append(
" Error column of all zeros indicates no error.\n")
return
#!/usr/bin/env python from plot_data.PlotHelper import PlotHelper import numpy as np import os xtest = np.arange(-10, 10, 1) ytest = np.sin(xtest) ytest2 = np.cos(xtest) xdatalist = [xtest, xtest] ydatalist = [ytest, ytest2] xerrlist = [None, None] yerrlist = [np.ones(len(ytest)) * 0.01, np.ones(len(ytest2)) * 0.02] kwargs = dict() kwargs['xdatalist'] = xdatalist kwargs['ydatalist'] = ydatalist kwargs['xerrlist'] = xerrlist kwargs['yerrlist'] = yerrlist kwargs['labellist'] = ['sine', 'cosine'] kwargs['marklargest'] = "2,3" kwargs['mlabellist'] = None kwargs['guideline'] = 1 kwargs['xlabel'] = "Number" kwargs['ylabel'] = "Function value" kwargs['plotlabel'] = "sine_cos_overlay" kwargs['save_path'] = os.path.join(os.getcwd(), "save_testing") kwargs['notelist'] = ["Test plot"] myph = PlotHelper(**kwargs) #myph.multiple_overlay() myph2 = PlotHelper() myph2.test_all()
def make_series_feature_plot(self, group=None):
"""Make series feature plot
"""
plabel = "feature_plot_group_%s" % str(group)
self.readme_list.append(" %s\n" % plabel)
pdict = dict()
group_notelist = list()
if not (self.plot_filter_out is None):
group_notelist.append("Data not displayed:")
for (feature, symbol, threshold) in self.plot_filter_out:
group_notelist.append(" %s %s %s" %
(feature, symbol, threshold))
testsets = list(self.sf_dict.keys())
testsets.sort()
for testset in testsets:
ts_sf_td = self.sf_dict[testset].testing_dataset
gfeat = self.training_dataset.grouping_feature
if group is None:
if ts_sf_td.target_data is None:
measured = None
else:
measured = ts_sf_td.target_data
feature_data = ts_sf_td.data[self.feature_plot_feature]
predicted = ts_sf_td.target_prediction
else:
if ts_sf_td.target_data is None:
measured = None
else:
measured = ts_sf_td.target_data[ts_sf_td.data[gfeat] ==
group]
feature_data = ts_sf_td.data[self.feature_plot_feature][
ts_sf_td.data[gfeat] == group]
predicted = ts_sf_td.target_prediction[ts_sf_td.data[gfeat] ==
group]
if measured is None:
pass
elif len(measured) == 0:
pass
else:
series_label = "%s measured" % testset
pdict[series_label] = dict()
pdict[series_label]['xdata'] = feature_data
pdict[series_label]['xerrdata'] = None
pdict[series_label]['ydata'] = measured
if len(predicted) == 0:
pass
else:
series_label = "%s predicted" % testset
pdict[series_label] = dict()
pdict[series_label]['xdata'] = feature_data
pdict[series_label]['xerrdata'] = None
pdict[series_label]['ydata'] = predicted
series_list = list(pdict.keys())
if len(series_list) == 0:
logging.info("No series for plot.")
return
addl_kwargs = dict()
addl_kwargs['guideline'] = 0
addl_kwargs['save_path'] = os.path.join(self.save_path, plabel)
addl_kwargs['xlabel'] = self.feature_plot_xlabel
addl_kwargs['ylabel'] = self.feature_plot_ylabel
addl_kwargs['markers'] = self.markers
addl_kwargs['outlines'] = self.outlines
addl_kwargs['linestyles'] = self.linestyles
addl_kwargs['legendloc'] = self.legendloc
addl_kwargs['sizes'] = self.sizes
faces = list()
for fidx in range(0, len(self.markers)):
faces.append("None")
addl_kwargs['faces'] = faces
addl_kwargs['group_dict'] = pdict
addl_kwargs['plotlabel'] = plabel
addl_kwargs['outlying_groups'] = series_list
addl_kwargs['notelist'] = list(group_notelist)
myph = PlotHelper(**addl_kwargs)
myph.plot_group_splits_with_outliers()
return