def plot_forest_features(self,
drop: bool = True,
norm: bool = False,
thresholds: list = None,
idx: int = -1):
"""
Plot the forest features obtained previously.
:return: --- <class 'NoneType'>
"""
if thresholds is None:
thresholds = [1]
else:
thresholds = sorted(thresholds)
max_features = ['sqrt', 'log2']
n_estimators = [5, 10, 25, 50, 75, 100, 150, 200, 250, 300]
thr_len = len(thresholds)
fig, axs = plt.subplots(thr_len, 2, figsize=(10, 4), squeeze=False)
for i in range(thr_len):
aux = self.compute_forest_features(drop, norm, thresholds[i], idx)
for k in range(len(max_features)):
layout = pF.LayoutStyleObject(
title='Random Forests with %s features' % max_features[k] +
'[T=' + str(thresholds[i]) + ']',
xlabel='Number of estimators',
ylabel='Accuracy',
grid=True)
pF.multiple_line_chart(axs[i, k],
n_estimators,
aux[k],
layout=layout,
percentage=True)
plt.show()
def plot_histogram(self, col_index: list = None, n_graphs: int = 5):
"""
Plots the histogram distribution for the given index columns, displaced in a matrix with n_graphs length.
:return: --- <class 'NoneType'>
"""
self.process_timer = time.time()
if col_index is None:
col_index = [2, 3, 4, 5, 6]
columns = (self.data.iloc[:, col_index]).columns
rows, cols = pF.choose_grid(len(columns) - 1, n_graphs)
fig, axs = plt.subplots(rows,
cols,
figsize=(cols * n_graphs, rows * n_graphs),
squeeze=False)
i, j = 0, 0
for n in range(len(columns)):
pF.histogram(axs[i, j],
self.data[columns[n]],
layout=pF.LayoutStyleObject(title='Histogram for %s' %
columns[n],
xlabel='',
ylabel='',
grid=True))
i, j = (i + 1, 0) if (n + 1) % cols == 0 else (i, j + 1)
plt.show()
self.get_timer("plot_histogram")
def plot_knn_var_threshold(self,
drop: bool = True,
norm: bool = False,
thresholds: list = None,
n_iter: int = 20,
parity: bool = False,
smote: bool = False):
"""
Plots the knn accuracy as function of n for given thresholds list.
:return: --- <class 'NoneType'>
"""
if thresholds is None:
thresholds = [
0.81, 0.83, 0.85, 0.87, 0.89, 0.91, 0.93, 0.95, 0.97, 0.99
]
else:
thresholds = sorted(thresholds)
if parity:
nvalues = [2 * i + 2 for i in range(round(n_iter / 2))]
else:
nvalues = [2 * i + 1 for i in range(round(n_iter / 2))]
if drop:
x = len(thresholds)
fig, axs = plt.subplots(int(x / 5),
int(x / (x / 5)),
figsize=(15, 6),
squeeze=False)
fig.subplots_adjust(hspace=.5, wspace=.001)
axs = axs.ravel()
for j in range(int(len(thresholds))):
axs[j].set_title('n')
layout = pF.LayoutStyleObject(
title='KNN variants - threshold ' + str(thresholds[j]),
xlabel='n',
ylabel='accuracy',
grid=True)
pF.multiple_line_chart(axs[j],
nvalues,
self.compute_knn(drop,
norm,
thresholds[j],
n_iter,
parity,
smote=smote),
layout=layout,
percentage=True,
legends=True)
fig.tight_layout()
plt.show()
else:
plt.figure()
pF.multiple_line_chart(plt.gca(),
nvalues,
self.compute_knn(drop, norm, thresholds),
title='KNN variants',
xlabel='n',
ylabel='accuracy',
percentage=True)
plt.show()
def plot_num_var_threshold(self,
tmin: float = 0,
tmax: float = 1,
step: float = 0.01):
"""
Plots the number of variables as a function of the threshold defined to drop.
Goes from 'tmin' to 'tmax' on 'step' steps.
:return: --- <class 'NoneType'>
"""
indexes = [0, 1, 2, 3, 4, 5, 6, 7]
plt.figure(figsize=(7, 5))
for i in indexes:
print(i)
vals = self.compute_num_var_threshold(tmin=tmin,
tmax=tmax,
step=step,
idx=i)
pF.single_line_chart(
plt.gca(),
vals[0],
vals[1],
layout=pF.LayoutStyleObject(
title='Number of Variables as a function of the Threshold',
xlabel='Threshold',
ylabel='Number of Variables',
grid=True),
plotstyle=pF.PlotStyleObject(color='Blue',
marker='o',
alpha=0.5))
plt.show()
def plot_tree_criteria(self,
drop: bool = True,
norm: bool = False,
thresholds: list = None,
idx: int = -1):
"""
For the given thresholds, plots the decision trees criteria, according to 'entropy' and 'gini' criteria.
:return: --- <class 'NoneType'>
"""
if thresholds is None:
thresholds = [1]
else:
thresholds = sorted(thresholds)
criteria = ['entropy', 'gini']
min_samples_leaf = [.10]
thr_len = len(thresholds)
fig, axs = plt.subplots(thr_len, 2, figsize=(10, 4), squeeze=False)
for i in range(thr_len):
for k in range(len(criteria)):
layout = pF.LayoutStyleObject(
title='Decision Trees with %s criteria' % criteria[k] +
'[T=' + str(thresholds[i]) + ']',
xlabel='Number of estimators',
ylabel='Accuracy',
grid=True)
pF.multiple_line_chart(axs[i, k],
min_samples_leaf,
self.compute_tree_criteria(
drop, norm, thresholds[i], idx)[k],
layout=layout,
percentage=True)
plt.show()
def draw_plots_none(var_x, var_y, colors):
aux_title = ''
for i in range(len(var_y)):
aux_title += var_y[i]
aux_title += '/'
aux_title = aux_title[:-1]
# define layout style object
layout = pF.LayoutStyleObject(aux_title + ' vs ' + var_x,
var_x,
aux_title,
grid=True)
# define first plot style object
n = 0
plot_style = {}
for i in range(len(var_y)):
plot_style[str(n)] = pF.PlotStyleObject(color=colors[n],
legend=var_y[i],
marker='o',
markersize=5,
linewidth=0)
n += 1
# define data to plot
n = 0
xvalues = {}
if var_x == 'ID':
for i in range(len(var_y)):
xvalues[str(n)] = data[var_y[i]].index.values
n += 1
else:
for i in range(len(var_y)):
xvalues[str(n)] = data[var_x].values
n += 1
n = 0
yvalues = {}
for i in range(len(var_y)):
yvalues[str(n)] = data[var_y[i]].values
n += 1
# define figure
plt.figure(figsize=(13, 6))
pF.multiple_plots(plt.gca(),
xvalues,
yvalues,
layout,
plot_style,
legends=True)
plt.show()
def plot_naive_bayes_var_threshold(self,
drop: bool = True,
norm: bool = False,
thresholds: list = None,
smote: bool = False):
"""
Plots the knn accuracy as function of n for given thresholds list.
:return: --- <class 'NoneType'>
"""
if thresholds is None:
thresholds = [
0.81, 0.83, 0.85, 0.87, 0.89, 0.91, 0.93, 0.95, 0.97, 0.99
]
else:
thresholds = sorted(thresholds)
if norm:
x_vals = ['Gaussian', 'Multinomial', 'Bernouly']
y_vals = [[], [], []]
else:
x_vals = ['Gaussian', 'Bernouly']
y_vals = [[], []]
for i in thresholds:
aux = self.compute_naive_bayes(drop, norm, i, smote=smote)
y_vals[0].append(aux[1][0])
y_vals[1].append(aux[1][1])
if norm:
y_vals[2].append(aux[1][2])
plt.figure()
layout = pF.LayoutStyleObject(
title='Naive Bayes accuracy as function of threshold',
xlabel='Threshold',
ylabel='Accuracy',
grid=True)
pF.multiple_bar_chart(plt.gca(),
thresholds,
y_vals,
x_vals,
layout=layout,
legends=True,
percentage=True)
plt.show()